JPH09259110A - Data variance estimation device and method therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To produce accurate data in a short time without preparing history data. SOLUTION: A change value/estimation data storage means 14 stores the change value between the current measured value and the next one and the estimation data in every area and step to which the current measured value belongs. An estimation accuracy calculation means 15 calculates the estimation accuracy based on the difference between the measured value and the estimated value. A change value/estimation accuracy detection means 17 detects the past change value and estimation accuracy corresponding to the area and step belonging to the current measured value. A change value calculation means 18 calculates the change value by performing the weighting to the past change value based on the estimation accuracy. Then a data estimation means 19 calculates the estimated value by adding the change value to the measured value.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data fluctuation prediction device and a data fluctuation prediction method for predicting future fluctuations of data that occur in time series.

[0002]

2. Description of the Related Art A conventional data fluctuation predicting apparatus will be described with reference to FIG.

FIG. 18 is a block diagram showing a system of a conventional data fluctuation predicting apparatus. The data detection means 100 detects current data. History data creating means 101
Stores, as history data, data that matches a predetermined standard from the detected data. Similarity calculation means 10
2 compares the current data detected by the data detecting means 100 with the past history data stored in the history data creating means 101 according to a predetermined reference similarity,
If there is past history data similar to the current data, this is detected as similar data. The data prediction unit 103 sets the similar data detected by the similarity calculation unit 102 as a future predicted value of the current variation data.

[0004]

In the above-mentioned conventional data fluctuation predicting apparatus, since it is necessary to save the data that meets a predetermined standard from the detected data as history data, it is attempted to improve the prediction accuracy. Then, there is a problem that the number of history data becomes huge.

Further, since the determination of the degree of similarity is performed by comparing predetermined criteria with each other, it takes a lot of time to retrieve past historical data similar to the current data, and at the same time, it is necessary to retrieve the current data. Since there is variation in the degree of similarity between the data and the historical data in the past, there is a problem that the prediction accuracy of the obtained similar data is poor.

The present invention solves the above problems all at once.
An object of the present invention is to provide a data fluctuation prediction device and a data fluctuation prediction method that can shorten the prediction time required for predicting data without providing historical data and greatly improve the prediction accuracy of prediction values.

[0007]

[Means for Solving the Problems] In order to achieve the above-mentioned object, the solution means taken by the invention of claim 1 is a data fluctuation prediction device for predicting future fluctuations of measured values observed in time series. A divided area creating means for creating a measured value divided area by dividing the range in which the measured values can be taken at predetermined intervals,
It is the difference between the first measurement value observed during the first observation and the second measurement value observed during the second observation after the first observation, for each of the measurement value division regions. A first change amount calculating unit that calculates a first change amount, and a prediction accuracy based on a difference between the predicted value at the time of the second measurement and the second measured value for each of the measured value divided regions. Prediction accuracy creating means to be created, and the first change amount calculated by the first change amount calculating means and the prediction accuracy created by the prediction accuracy creating means are stored for each of the measurement value divided areas. From the data storage means and the first change amount and the prediction accuracy stored in the data storage means, the first change amount and the prediction accuracy belonging to the measurement value divided region to which the measured value observed this time belongs are calculated. The change amount / prediction accuracy detection means for detecting the change amount / prediction accuracy Second change amount calculation means for calculating a second change amount by weighting the first change amount detected by the means with the prediction accuracy detected by the change amount / prediction accuracy detection means; It is configured to include a predicted value creating unit that creates a predicted value for the next observation by adding the second amount of change calculated by the second amount of change calculating unit to the observed measurement value. .

According to the structure of claim 1, the difference between the first measurement value observed during the first observation and the second measurement value observed during the second observation, which is calculated for each measurement value divided area, is calculated. The second variation is performed by weighting the certain first variation amount by the prediction accuracy created based on the difference between the second measurement value and the second measurement value for each measurement value division area. After creating the amount, the second variation is added to the measured value observed this time to create the predicted value for the next observation, so the obtained predicted value is weighted by the prediction accuracy for each measurement value divided region. ing.

The solution means taken by the invention of claim 2 is intended for a data fluctuation predicting device for predicting future fluctuations of measured values observed in a time series, and the number of observed objects or the number of observations appearing in the time series is determined. Division area creating means for creating an observation number divided area by dividing at every predetermined interval, and for each observation number divided area,
A first amount of change, which is the difference between the first measurement value observed during the first observation and the second measurement value observed during the second observation after the first observation, is calculated. A first change amount calculating means; a prediction accuracy creating means for creating a prediction accuracy based on a difference between the predicted value at the time of the second measurement and the second measured value for each of the observation number division regions; Data storage means for storing the first change amount calculated by the first change amount calculation means and the prediction accuracy created by the prediction accuracy creation means for each of the observation number division regions, and the data storage. Change amount / prediction accuracy detection for detecting the first change amount and prediction accuracy belonging to the observation number division region to which the measurement value observed this time belongs, from the first change amount and prediction accuracy stored in the means. Means and the first amount detected by the change amount / prediction accuracy detection means. The second variation amount calculating means for calculating the second variation amount by weighting the variation amount of the above with the prediction accuracy detected by the variation amount / prediction accuracy detecting means, and the measurement value observed this time as described above. The second change amount calculating means is added to the second change amount to generate a predicted value for the next observation.

According to the structure of claim 2, the difference between the first measurement value observed during the first observation and the second measurement value observed during the second observation, which is calculated for each observation number divided region, is calculated. The second variation is performed by weighting the certain first variation amount with the prediction accuracy created based on the difference between the second measurement value and the second measurement value for each observation number division region. After creating the quantity, the second variation is added to the measurement value observed this time to create the prediction value for the next observation. Therefore, the obtained prediction value is weighted by the prediction accuracy for each observation number division region. ing.

The solution means taken by the invention of claim 3 is intended for a data fluctuation predicting device for predicting future fluctuations of measured values observed in a time series, and a first fluctuation observed at the first observation.
Change rate step creation in which the change rate step is created by dividing the rate of change between the measured value of 1 and the second measured value observed in the second observation after the first observation into predetermined rates. Means and
First change amount calculation means for calculating a first change amount which is a difference between the first measured value and the second measured value for each of the change rate steps; and for each of the change rate steps, Prediction accuracy creating means for creating prediction accuracy based on the difference between the predicted value at the second measurement and the second measured value; and the first change amount calculating means for each of the change rate stages. The first change amount and the prediction accuracy created by the prediction accuracy creating unit, and the data storage unit that stores the prediction accuracy, and the first change amount and the prediction accuracy stored in the data storage unit.
A change amount / prediction accuracy detection means for detecting a first change amount and prediction accuracy belonging to the change rate stage to which the measured value observed this time belongs, and a first change detected by the change amount / prediction accuracy detection means. A second change amount calculation means for calculating a second change amount by weighting the amount with the prediction accuracy detected by the change amount / prediction accuracy detection means, and a second change amount for the measured value observed this time. And a predicted value creating means for creating a predicted value at the time of the next observation by adding the second amount of change calculated by the changing quantity calculating means.

According to the structure of claim 3, the first measurement value and the second measurement value which are calculated at each change rate stage and are observed at the first observation time.
Based on the difference between the predicted value at the second measurement and the second measured value for each change rate step, for the first amount of change that is the difference from the second measured value observed at Prediction value obtained by adding the second variation to the measurement value observed this time to create the prediction value for the next observation after creating the second variation by weighting with the created prediction accuracy Is weighted according to the prediction accuracy for each change rate step.

The solution means taken by the invention of claim 4 is intended for a data fluctuation predicting device for predicting future fluctuations of measured values observed in a time series, and from the measured value observed at the start of observation to every observation time. Change amount step creating means for creating a change amount step by dividing the total sum of the change amounts up to each observed value into predetermined steps, and each change amount step is observed at the first observation time. A first change amount calculating means for calculating a first change amount, which is a difference between a first measurement value and a second measurement value observed in a second observation after the first observation. , The predicted value at the time of the second measurement and the second
Prediction accuracy creating means for creating a prediction accuracy based on the difference between the measured value and the first change amount calculated by the first change amount calculating means and the prediction accuracy creating means for each of the change amount stages. From the data storage means for storing the prediction accuracy created by the above, and the first change amount and the prediction accuracy stored in the data storage means, to the change amount stage to which the measured value observed this time belongs. A change amount / prediction accuracy detection unit that detects the first change amount and prediction accuracy that belong, and a change amount / prediction accuracy detection unit for the first change amount that is detected by the change amount / prediction accuracy detection unit. Second change amount calculating means for calculating the second change amount by weighting according to the detected prediction accuracy, and second change calculated by the second change amount calculating means for the measured value observed this time. Next add the amount It is an arrangement and a prediction value generating means for generating a prediction value at the time observed.

According to the structure of claim 4, the first measurement value and the second measurement value which are calculated for each change amount stage and are observed at the first observation time.
Based on the difference between the predicted value at the second measurement and the second measured value for each change amount stage, for the first change amount that is the difference from the second measured value observed at Prediction value obtained by adding the second variation to the measurement value observed this time to create the prediction value for the next observation after creating the second variation by weighting with the created prediction accuracy Is weighted according to the prediction accuracy for each change amount stage.

The solution means taken by the invention of claim 5 is directed to a data fluctuation predicting device for predicting future fluctuations of measured values observed in a time series, and the range of the measured values is divided at predetermined intervals. And a first divided area creating means for creating a measured value divided area and a second divided area creating means for creating an observed number divided area by dividing the number of observed objects or the number of observations appearing in time series at predetermined intervals. Means, and the rate of change between the first measurement value observed during the first observation and the second measurement value observed during the second observation after the first observation step for each predetermined ratio A rate-of-change step creating means that divides and creates rate-of-change steps, and the total amount of change from the measured value observed at the start of observation to each measured value observed at each observation Change amount stage creating means for creating a change amount stage,
Measurement value division area, observation number division area, measurement value division area change amount, observation number division, which is the difference between the first measurement value and the second measurement value for each change rate step and change amount step A first change amount calculating means for calculating a region change amount, a change rate step change amount, and a change amount step change amount, respectively, and the measurement value division area, the observation number division area, the change rate step and the change amount step, Based on the difference between the predicted value at the time of the second measurement and the second measured value, the measurement value division region prediction accuracy, the observation number division region prediction accuracy, the change rate step prediction accuracy, and the change amount step prediction accuracy are created, respectively. Prediction accuracy creating means, the measurement value division area, the observation number division area, the measurement value division area change amount and the observation number division calculated by the first change amount calculation means for each change rate step and each change amount step Area change amount, change rate step change amount, and change amount step change Quantity and the measurement value division area prediction accuracy created by the prediction accuracy creation means, the observation number division area prediction accuracy, the change rate step prediction accuracy, and the change amount step prediction accuracy, and the data storage means. From the measured value divided area change amount and the measured value divided area prediction accuracy stored in the means, the measured value divided area change amount and measured value divided area prediction belonging to the measured value divided area to which the measured value observed this time belongs The detection that outputs the accuracy, the observation number division region change amount and the observation number division region prediction accuracy stored in the data storage means, the observation that belongs to the observation number division region to which the measurement value observed this time belongs The number of divided regions change amount and the number of observed regions divided region prediction accuracy are detected and output, and the change rate step change amount and the change rate step prediction accuracy stored in the data storage means are calculated. From this, the change rate step change amount belonging to the change rate step to which the measurement value observed this time belongs and the change rate step prediction accuracy are detected and output, and the change rate step change amount and change stored in the data storage means. A change amount / prediction accuracy detection means for detecting and outputting the change amount step change amount and the change amount step prediction accuracy belonging to the change amount step to which the measurement value observed this time belongs from the quantity step prediction accuracy, and the change. For the measurement value divided area change amount, the number of observations divided area change amount, the change rate step change amount and the change amount step change amount output from the quantity / prediction accuracy detection means,
A second amount for calculating a variation amount by weighting the measurement value division region prediction precision, the observation number division region prediction precision, the change rate stage prediction precision, and the change amount stage prediction precision output from the change amount / prediction precision detection means. A configuration comprising a change amount calculation means and a predicted value generation means for adding a change amount calculated by the second change amount calculation means to the measured value observed this time to create a predicted value for the next observation; To do.

According to the structure of claim 5, a measurement value division area,
Observed number division area, variation rate step and variation amount Measured value division area variation amount, observation number division area variation amount calculated based on the difference between the predicted value and the second measurement value at the time of the second measurement , The change rate step change amount and the change amount step change amount, the measurement value division area, the observation number division area, the predicted value at the second measurement and the second measurement value for each of the change rate step and the change amount step After creating the amount of change by weighting with the prediction accuracy created based on the difference of, the above-mentioned amount of change is added to the measured value observed this time to create the predicted value for the next observation The values are weighted according to the prediction accuracy for each of the measurement value division area, the observation number division area, the change rate step, and the change amount step.

According to a sixth aspect of the present invention, in the configuration of the first to fifth aspects, the predicted value at the time of the second measurement in the prediction accuracy creating means is a predicted value created by the predicted value creating means. It is something to add.

According to a seventh aspect of the present invention, in the configuration according to the fifth or sixth aspect, the change amount / prediction accuracy detecting means includes the measurement value division area change amount and the measurement value division area prediction stored in the data storage means. After detecting the measurement value division area change amount and measurement value division area prediction accuracy that belong to the measurement value division area to which the measurement value observed this time belongs from the accuracy, then detect the measurement value division area change amount and measurement value division The average value of the region prediction accuracy is calculated and output, and from the observation number divided region change amount and the observation number divided region prediction accuracy stored in the data storage means, the observation to which the measurement value observed this time belongs After detecting the number of observations divided region change amount and the number of observations divided region prediction accuracy that belong to the number division region, each average value of the detected number of observations divided region change and the number of observations divided region prediction accuracy is output, After detecting the change rate step change amount and the change rate step prediction accuracy belonging to the change rate step to which the measurement value observed this time belongs from the change rate step change amount and the change rate step prediction accuracy stored in the means. , The average value of the detected change rate step change amount and change rate step prediction accuracy is output, and this time is observed from among the change step change amount and the change step prediction accuracy stored in the data storage means. The change amount step change amount belonging to the change amount stage to which the measurement value belongs and the change amount stage prediction accuracy are detected, and each average value of the detected change amount stage change amount and the change amount stage prediction accuracy is output. The change amount calculation means, for each average value of the measurement value division region change amount, the observation number division region change amount, the change rate step change amount and the change amount step change amount output from the change amount / prediction accuracy detection means. , The change A configuration in which the change amount is calculated by weighting the measurement value divided region prediction precision, the observation number divided region prediction precision, the change rate stage predicted precision, and the change amount stage predicted precision output from the prediction precision detection means. Is added.

According to an eighth aspect of the present invention, in the configuration of the fifth to seventh aspects, the prediction precision creating means sets the prediction precision = for each of the measurement value division region, the observation number division region, the change rate stage and the change amount stage. 1 / | predicted value at the time of the second measurement−second measured value | by the formula, the measurement value division region prediction accuracy, the observation number division region prediction accuracy, the change rate step prediction accuracy, and the change amount step prediction accuracy The configuration for creating each is added.

According to a ninth aspect of the present invention, in the structure according to the fifth to eighth aspects, the second change amount calculating means calculates the change amount as follows: change amount = measured value divided area changed amount × measured value divided area Prediction accuracy / (measurement value division area prediction accuracy + observation number division area prediction accuracy + change rate step prediction accuracy + change amount step prediction accuracy) + observation number division area change quantity × observation number division area prediction accuracy / (Measurement value divided area prediction accuracy + observation number divided area prediction accuracy + change rate step prediction accuracy + change amount step prediction accuracy) + change rate step change amount × change rate step prediction accuracy / (measurement Value-division region prediction accuracy + number of observations region-division region prediction accuracy + change rate step prediction accuracy + change amount step prediction accuracy) + change amount step change amount x change amount step prediction accuracy / (measurement value divided region prediction Accuracy + number of observations It is intended to add the configuration for calculating the expression region prediction accuracy + the rate of change stage prediction accuracy + the variation stage prediction accuracy).

According to a tenth aspect of the present invention, in the configuration according to the fifth to eighth aspects, the second change amount calculating means changes the change amount into the measured value divided area changed amount, the observed number divided area changed amount, and the changed number. Of the rate step change amount and the change amount step change amount, the one that corresponds to the largest prediction accuracy among the measurement value divided region prediction accuracy, the observation number divided region prediction accuracy, the change rate step predicted accuracy, and the change amount step predicted accuracy. A configuration for selecting and calculating is added.

The means for solving the problems of the eleventh aspect of the present invention is intended for a data fluctuation predicting device for predicting future fluctuations of measured values observed in a time series, and the range of the measured values is divided at predetermined intervals. And a first divided area creating means for creating a measured value divided area and a second divided area creating means for creating an observed number divided area by dividing the number of observed objects or the number of observations appearing in time series at predetermined intervals. Means, and the rate of change between the first measurement value observed during the first observation and the second measurement value observed during the second observation after the first observation step for each predetermined ratio A rate-of-change step creating means that divides and creates rate-of-change steps, and the total amount of change from the measured value observed at the start of observation to each measured value observed at each observation Change amount stage creating means for creating a change amount stage,
Measurement value division area, observation number division area, measurement value division area change amount, observation number division, which is the difference between the first measurement value and the second measurement value for each change rate step and change amount step This time, from the data storage means for calculating and storing the area change amount, the change rate step change amount, and the change amount step change amount, and the measured value divided area change amount stored in the data storage means The first change amount is detected as a first change amount that belongs to the measurement value divided region to which the measured value belongs and is frequently detected, and is detected this time from the observation number divided region change amount stored in the data storage unit. The second change amount that has a high frequency and belongs to the observation number division region to which the measured value belongs is detected as the second change amount, and the currently observed measurement is made from the change rate step change amounts stored in the data storage means. The variable to which the value belongs A third change amount that belongs to the rate stage and is frequently detected is detected, and the change amount stage change amount stored in the data storage means is changed to the change amount stage to which the measured value observed this time belongs. Change amount detection means for detecting a high-frequency one that belongs and has a high frequency,
An average change amount calculation unit that calculates an average change amount by averaging the first change amount, the second change amount, the third change amount, and the fourth change amount detected by the change amount detection unit; The present invention is configured to include a predicted value creation unit that creates the predicted value for the next observation by adding the average change amount calculated by the average change amount calculation unit to the measured value observed this time.

According to the eleventh aspect, the first variation that belongs to the measurement value division region to which the measurement value observed this time belongs and has a high frequency, belongs to the observation number division region to which the measurement value observed this time belongs, and frequency. The second change amount is high, and the third change amount is high and belongs to the change rate stage to which the measured value observed this time belongs.
The average change amount is calculated by averaging the change amount of 4 and the 4th change amount that belongs to the change amount stage to which the measured value observed this time belongs and has a high frequency, and then calculates the average change amount to the measured value observed this time. In addition, in order to create a predicted value for the next observation, the predicted value obtained is calculated based on four variable factors: the measurement value division area, the observation number division area, the change rate step and the change amount step.

The means for solving the problems of the twelfth aspect of the invention is intended for a data fluctuation predicting device for predicting future fluctuations of measured values observed in a time series, and the range of the measured values is divided at predetermined intervals. And a first divided area creating means for creating a measured value divided area and a second divided area creating means for creating an observed number divided area by dividing the number of observed objects or the number of observations appearing in time series at predetermined intervals. Means, and the rate of change between the first measurement value observed during the first observation and the second measurement value observed during the second observation after the first observation step for each predetermined ratio A rate-of-change step creating means that divides and creates rate-of-change steps, and the total amount of change from the measured value observed at the start of observation to each measured value observed at each observation Change amount stage creating means for creating a change amount stage,
Measurement value division area, observation number division area, change rate step and change amount step, which is the difference between the first measurement value and the second measurement value. The data storage means for calculating and storing the area change amount, the change rate step change amount, and the change amount step change amount respectively, and the measured value division area change amount stored in the data storage means, which was observed this time. The first change amount is calculated by averaging the change amounts of the measurement value divided regions belonging to the measurement value divided region to which the measurement value belongs, and the first change amount is calculated, and the first change amount is observed among the observation number divided region change amounts stored in the data storage means. The second change amount is calculated by averaging the change amounts of the observation number division regions belonging to the observation number division region to which the measured value belongs, and is observed this time among the change rate step change amounts stored in the data storage means. The variable to which the measured value belongs The change rate step change amount belonging to the rate step is averaged to calculate a third change rate, and the change rate step to which the currently observed measurement value belongs among the change rate step change rates stored in the data storage means. Change amount calculation means for calculating the fourth change amount by averaging the change amount step change amounts belonging to, and the first change amount, the second change amount, and the third change amount calculated by the change amount calculation means. Amount and the fourth change amount are averaged to calculate the average change amount, and the average change amount calculated by the average change amount calculation unit is added to the measured value observed this time, and the next observation time is obtained. And a predictive value creating means for creating a predicted value of.

According to the twelfth aspect, the first change amount, which is the average value of the change amount of the measurement value division areas belonging to the measurement value division area to which the measurement value observed this time belongs, and the observation to which the measurement value observed this time belongs The second change amount, which is the average value of the change amount of the observation value division region belonging to the number-division region, and the third change amount, which is the average value of the change rate stage change amount of the change rate stage to which the measured value observed this time belongs. , And the fourth change amount, which is the average value of the change amount step change amounts belonging to the change amount stage to which the measured value observed this time belongs, is averaged to calculate the average change amount, and then averaged to the measured value observed this time. Since the predicted value for the next observation is created by adding the amount of change, the predicted value obtained is the measured value division region, the observation number division region, the change rate stage, and the change amount stage.
It is calculated based on two fluctuation factors.

The means for solving the problems according to the thirteenth aspect of the present invention is directed to a data fluctuation prediction method for predicting future fluctuations of measured values observed in a time series, and the range of the measured values is divided at predetermined intervals. To create a measurement value division area, and divide the number of observations or the number of observations that appear in time series at predetermined intervals to create an observation number division area, and the first measurement value observed at the first observation And the rate of change between the second measurement value observed during the second observation after the first observation and the second measurement value after the first observation are divided into steps at a predetermined rate to create a rate of change step, which is observed at the start of observation. Area / step creation process that creates a change amount step by dividing the total sum of the change amount from the measurement value to each measurement value observed at each observation step by a predetermined ratio, the measurement value division area, the observation number division For each region, change rate step and change amount step, the first
Difference between the measured value and the second measured value, the measured value divided area change amount, the observed number divided area change amount, the change rate step change amount, and the change amount step change amount A calculation step and a measurement value based on the difference between the predicted value at the second measurement and the second measurement value for each of the measurement value division area, the observation number division area, the change rate step and the change amount step. Divided area prediction accuracy, number of observations divided area prediction accuracy, prediction accuracy creating step of creating change rate step prediction accuracy and change step prediction accuracy, respectively, from the measurement value split area change amount,
The first change amount that is frequently and frequently belongs to the measurement value division region to which the measurement value observed this time belongs is detected, and the measurement value observed this time belongs to the change amount of the observation number division region. Those that belong to the number-of-observations division region and have a high frequency are detected as the second change amount, and from the change rate stage change amount, those that belong to the change rate stage to which the measurement value observed this time belongs and have a high frequency Change amount that is detected as a third change amount, and that among the change amount step change amounts, the one that belongs to the change amount step to which the measurement value observed this time belongs and has a high frequency is detected as the fourth change amount. Detection step, the measurement value divided area prediction accuracy, the number of observations divided area prediction accuracy, and the change rate step for the first change amount, the second change amount, the third change amount, and the fourth change amount. For prediction accuracy and change level prediction accuracy And a second change amount calculating step of calculating a change amount by weighting, and a predicted value creating step of adding the change amount to the measured value observed this time to create a predicted value for the next observation. It is to be configured.

According to the configuration of claim 13, measurement is performed for each change amount that belongs to the measurement value division region to which the measurement value observed this time belongs, the observation number division region, the change rate stage and the change amount stage, and has a high frequency. After creating the amount of change by weighting with the prediction accuracy of the value division area, the observation number division area, the change rate step and the change amount step, add the above-mentioned change amount to the measurement value observed this time, and In order to create the predicted value, the obtained predicted value is calculated based on four variable factors of the measurement value division area, the observation number division area, the change rate step and the change amount step, and is weighted by the prediction accuracy. There is.

The solution means taken by the invention of claim 14 is based on a data fluctuation prediction method for predicting future fluctuations of measured values observed in a time series, and the range of the measured values is divided into predetermined intervals. To create a measurement value division area, and divide the number of observations or the number of observations that appear in time series at predetermined intervals to create an observation number division area, and the first measurement value observed at the first observation And the rate of change between the second measurement value observed during the second observation after the first observation and the second measurement value after the first observation are divided into steps at a predetermined rate to create a rate of change step, which is observed at the start of observation. Area / step creation process that creates a change amount step by dividing the total sum of the change amount from the measurement value to each measurement value observed at each observation step by a predetermined ratio, the measurement value division area, the observation number division For each region, change rate step and change amount step, the first
Difference between the measured value and the second measured value, the measured value divided area change amount, the observed number divided area change amount, the change rate step change amount, and the change amount step change amount A calculation step and a measurement value based on the difference between the predicted value at the second measurement and the second measurement value for each of the measurement value division area, the observation number division area, the change rate step and the change amount step. Prediction accuracy creating step for creating the divided area prediction accuracy, the number of observations divided area prediction accuracy, the change rate step prediction accuracy, and the change step prediction accuracy, and the measurement value observed this time among the measurement value divided area change quantity Calculating a first variation amount by averaging the variation amounts of the measurement value division regions belonging to the measurement value division regions,
Of the observation number division region variation amounts, the observation amount division region variation amounts belonging to the observation number division region to which the measurement value observed this time belongs are averaged to calculate a second variation amount. The third change amount is calculated by averaging the change rate step change amounts belonging to the change rate stage to which the measured value observed this time belongs, and the measured value observed this time belongs to the change amount step change amount. An average change amount calculation step of calculating a fourth change amount by averaging the change amount step change amounts belonging to the change amount step, and the first change amount, the second change amount, the third change amount, and the fourth change amount. Change amount calculation step for calculating the change amount by weighting the change amount of the measurement value divided region prediction accuracy, the observation number divided region prediction accuracy, the change rate step prediction accuracy, and the change amount step prediction accuracy. And the amount of change in the measured value observed this time It is an arrangement and a Ete predicted value generation step of generating a predicted value of the next observation time.

According to the structure of claim 14, the measured value is divided into the measured value divided area to which the measured value observed this time belongs, the observed number divided area, and the average value of each change amount belonging to each of the change rate stage and the change amount stage. Prediction at the next observation by adding the above-mentioned amount of change to the measured value observed this time after creating the amount of change by weighting the divided regions, the number of observed regions, the rate of change and the amount of change level prediction To create a value
The obtained predicted values are the measurement value division area, the observation number division area,
It is calculated based on four variation factors of the change rate stage and the change amount stage, and is weighted by the prediction accuracy.

The invention of claim 15 is the invention of claim 13 or 14
In the configuration of, the prediction accuracy creating step, for each of the measurement value division area, the observation number division area, the change rate step and the change amount step,
Prediction accuracy = 1 / | predicted value at the time of the second measurement−the second
The measurement value | of the measurement value |
The number of observations divided region prediction accuracy, the change rate step prediction accuracy, and the change amount step prediction accuracy are respectively added.

According to a sixteenth aspect of the present invention, in the structure according to the thirteenth to fifteenth aspects, the second change amount calculating step calculates the change amount as follows: change amount = the first change amount × the measured value divided area prediction Accuracy / (measurement value division area prediction accuracy + observation number division area prediction accuracy + change rate step prediction accuracy + change amount step prediction accuracy) + the second change amount × observation number division area prediction accuracy / ( The measurement value division area prediction accuracy + the observation number division area prediction accuracy + the change rate step prediction accuracy + the change amount step prediction accuracy) + the third change quantity × the change rate step prediction accuracy / (the measurement value division Region prediction accuracy + the number of observations divided region prediction accuracy + the change rate step prediction accuracy + the change amount step prediction accuracy) + the fourth change amount × the change amount step prediction accuracy / (the measurement value divided region prediction accuracy + The number of observations divided region prediction accuracy + previous It is intended to add a configuration of calculating the equation of the rate of change stage prediction accuracy + the variation stage prediction accuracy).

According to a seventeenth aspect of the present invention, in the structure of the thirteenth to fifteenth aspects, the second change amount calculation step is the change amount calculation step.
Of the first change amount, the second change amount, the third change amount, and the fourth change amount, the measurement value divided region prediction accuracy, the observed number divided region prediction accuracy, the change rate step prediction accuracy, and the change amount. A configuration for selecting and calculating the one corresponding to the highest prediction accuracy among the step prediction accuracy is added.

A solution means taken by the invention of claim 18 is directed to a data fluctuation prediction method for predicting future fluctuations of measured values observed in a time series, and the range of the measured values is divided into predetermined intervals. To create a measurement value division area, and divide the number of observations or the number of observations that appear in time series at predetermined intervals to create an observation number division area, and the first measurement value observed at the first observation And the rate of change between the second measurement value observed during the second observation after the first observation and the second measurement value after the first observation are divided into steps at a predetermined rate to create a rate of change step, which is observed at the start of observation. Area / step creation process that creates a change amount step by dividing the total sum of the change amount from the measurement value to each measurement value observed at each observation step by a predetermined ratio, the measurement value division area, the observation number division For each region, change rate step and change amount step, the first
And a change amount calculating step of calculating a change amount of the measurement value divided region, a change amount of the observation number divided region, a change rate step change amount, and a change amount step change amount, which are differences between the measurement value of the measurement value and the second measurement value. Of the change amount of the measurement value divided region, the one that belongs to the measurement value divided region to which the measurement value observed this time belongs and has a high frequency is detected as the first change amount, and the change amount of the observation number divided region From among these, the one that belongs to the observation number division region to which the measurement value observed this time belongs and has a high frequency is detected as the second change amount, and from the change rate step change amount, the measurement value observed this time is A third change amount that belongs to the change rate stage and that has a high frequency is detected, and the change amount stage change amount belongs to the change amount stage to which the measurement value observed this time belongs and has a high frequency. The thing as the fourth variation A change amount detecting step of detecting Te, the first change amount and the second variation amount, a third variation and a fourth
The average change amount calculating means for averaging the change amounts of the above to calculate the average change amount, and a predicted value creation step of adding the average change amount to the measurement value observed this time to create a predicted value for the next observation The configuration is as follows.

According to the structure of claim 18, for each of the measurement value division area, the observation number division area, the change rate step and the change amount step,
A first change amount and a second change amount, which are calculated based on the difference between the predicted value at the second measurement and the second measured value and have a high frequency,
After calculating the average change amount by averaging the third change amount and the fourth change amount, the above-mentioned average change amount is added to the measurement value observed this time to create the predicted value for the next observation. The predicted value to be calculated is calculated on the basis of the four fluctuation factors of the measurement value division area, the observation number division area, the change rate stage and the change amount stage.

The solution means taken by the invention of claim 19 is directed to a data fluctuation prediction method for predicting future fluctuations of measured values observed in a time series, and the range of the measured values is divided at predetermined intervals. To create a measurement value division area, and divide the number of observations or the number of observations that appear in time series at predetermined intervals to create an observation number division area, and the first measurement value observed at the first observation And the rate of change between the second measurement value observed during the second observation after the first observation and the second measurement value after the first observation are divided into steps at a predetermined rate to create a rate of change step, which is observed at the start of observation. Area / step creation process that creates a change amount step by dividing the total sum of the change amount from the measurement value to each measurement value observed at each observation step by a predetermined ratio, the measurement value division area, the observation number division For each region, change rate step and change amount step, the first
Difference between the measured value and the second measured value, the measured value divided area change amount, the observed number divided area change amount, the change rate step change amount, and the change amount step change amount In the calculation step, the first change amount is calculated by averaging the change amounts of the measurement value division regions belonging to the measurement value division region to which the measurement value observed this time belongs among the measurement value division regions change amount, and the data storage The second change amount is calculated by averaging the change amounts of the observation number division regions belonging to the observation number division region to which the measurement value observed this time belongs among the change amounts of the observation number division regions stored in the means. The change rate step change amount belonging to the change rate step to which the measured value observed this time belongs among the change rate step change amounts stored in the storage means is averaged to obtain a third value.
Of the change amount step change amount stored in the data storage means, and the change amount step change amount belonging to the change amount step to which the measurement value observed this time belongs is averaged to obtain a fourth change. A second change amount calculating step for calculating an amount, and an average change amount for calculating an average change amount by averaging the first change amount, the second change amount, the third change amount, and the fourth change amount. The configuration includes a calculation step and a predicted value creation step of creating a predicted value for the next observation by adding the average change amount to the measured value observed this time.

According to the configuration of claim 19, the average value of the measured values divided area, the observed number divided area, the change rate stage, and the change amount belonging to each change amount stage are averaged to create an average change amount. After that, the average change amount is added to the measurement value observed this time to create the prediction value for the next observation, so the obtained prediction value is divided into the measurement value division region, the observation number division region, the change rate stage, and the change amount stage. It is calculated based on four variables.

[0037]

BEST MODE FOR CARRYING OUT THE INVENTION

(First Embodiment) A data fluctuation predicting apparatus according to a first embodiment of the present invention will be described below with reference to FIGS. 1 and 9 to 13. The first embodiment is a data fluctuation predicting apparatus that predicts fluctuations in the number of particles generated from a semiconductor manufacturing apparatus using a computer.

FIG. 1 shows a system block diagram of a data fluctuation predicting apparatus according to the first embodiment, and FIG. 9 shows an example of apparatus particle data managed in an XY coordinate system.
FIG. 10 shows a Y area history data management table, and FIG.
Shows an X area history data management table, FIG. 12 shows a change rate data management table, and FIG. 13 shows a change amount total data management table.

As shown in FIG. 1, the data variation predicting apparatus according to the first embodiment has a Y area determining means 10, an X area determining means 11, a change rate step dividing means 12, a change amount step dividing means 13, and a change. The amount / prediction data storage unit 14, the prediction accuracy calculation unit 15, the storage device 16, the change amount / prediction accuracy detection unit 17, the change amount calculation unit 18, and the data prediction unit 19 are managed in the XY coordinate system in time series. This is a device that predicts future changes in observed particle data. In addition, in the particle data shown in FIG. 9, the X-axis represents the elapsed time or the number of observations by replacing the number of processed wafers with each other, and the Y-axis represents the number of particles per 100 processed wafers (each observation). There is.

The Y region determining means 10 divides the range in which the number of particles detected by the semiconductor process device can be taken in the Y-axis direction at predetermined intervals. That is, the range from the minimum value to the upper limit value (maximum value) of the number of particles is set as the Y area, the Y area is divided into 6 equal parts, and the Y area number, which is a number for identification, is assigned to each of the divided areas. Give. still,
In the first embodiment, the Y region is divided into six equal parts, but the number of divisions is not limited, and may be divided into an appropriate number in consideration of the maximum value of the number of particles, the prediction accuracy, and the like.

The X area determining means 11 defines an X area as a range in which the number of processed wafers detected by the semiconductor processing apparatus can be taken in the X axis direction, and divides the X area at predetermined intervals, for example, every 100 wafers. Then, each divided area is given an X area number, which is a number for identification. The first
In the embodiment, the X region is divided in units of 100 wafers, but the unit of division is not limited, and the division unit is considered in consideration of the maximum value of the number of processed wafers in the past, the particle count measurement frequency, the prediction accuracy, and the like. Can be determined.

The rate-of-change dividing means 12 divides the rate of change of the number of particles with respect to the number of processed wafers detected by the semiconductor processing apparatus into predetermined steps (intervals). Specifically, (the number of particles at the time of this detection-the number of particles at the time of the previous detection) / (the number of processed particles at the time of this detection-
The absolute value of the change rate (the number of processed sheets at the time of the previous detection) (theoretically exists from 0 to infinity) is divided into, for example, 16 stages at a predetermined ratio, and the divided stages are used for identification. The change rate step number, which is the number of. In the first embodiment, the change rate is divided into 16 stages, but the unit of division (the number of stages) is not limited, and the number of processed wafers,
The division unit can be determined in consideration of the particle count measurement frequency or the prediction accuracy.

The variation amount grading means 13 graduates the total amount of variation in the number of particles with respect to the number of processed wafers detected by the semiconductor process apparatus into predetermined amounts. Specifically, the total amount of change in the number of particles is divided into stages, for example, every 60 particles, and a change amount step number, which is a number for identification, is given to each of the divided steps.
In addition, in the first embodiment, the total sum of the amounts of change is divided into stages by the number of particles 60, but the unit of dividing the stages is not limited, and the maximum value of the total sum of the past numbers of particles, the particle count measurement frequency, or the prediction The division unit can be determined in consideration of accuracy and the like. Further, the sum of change amounts may be the sum of absolute values of the respective change amounts, or the sum of relative values in consideration of + -of the change amounts.

The order of performing the processes performed by the Y area determining means 10, the X area determining means 11, the change rate step dividing means 12 and the change amount step dividing means 13 is not limited.

Next, the variation / prediction data storage means 14
Performs the following four processes when a certain measured value is observed. The order of performing this process is not particularly limited.

FIG. 10 shows the Y area history data management table on the auxiliary storage device of the computer having the Y area number, the current measured value, the change amount up to the next time, the next predicted value and the prediction accuracy as items. . Here, the current measurement value is the measurement value that was observed this time, and the change amount up to the next is the measurement value that was actually observed from the time of the next observation to the measurement value of the previous observation (corresponding to the measurement value that was observed this time) , The next predicted value is a predicted value predicted by the data prediction means 19 described later, and the prediction accuracy is the prediction accuracy calculated by the prediction accuracy calculation means 15 described later. Y at the time of this observation
Register the Y area number, the current measurement value and the next predicted value in the record of the Y area to which the measurement value observed this time belongs in the area history data management table, and at the time of the next observation, calculate the change amount up to the next time. Then, it is registered in the record of the Y area.

FIG. 11 shows the X area history data management table on the auxiliary storage device of the computer having the X area number, the current measured value, the amount of change until the next time, the next predicted value and the prediction accuracy as items. . The meaning of each item is the same as that of the Y area history data management table of FIG.
At the time of this observation, the X area number, the current measurement value, and the next predicted value are registered in the record of the X area to which the observed measurement value belongs in the X area history data management table. Change amount is calculated and registered in the record of the X area.

FIG. 12 shows a change rate data management table on the auxiliary storage device of the computer having the change rate stage, the current measured value, the next change amount, the next predicted value and the prediction accuracy as items. Figure 1 shows the meaning of each item
This is the same as the Y area history data management table of 0. At the time of this observation, the rate of change of the measured value observed this time from the time of the previous observation is calculated, and the change rate stage number and the current measured value are recorded in the record of the rate of change to which the changed rate calculated this time belongs in the data management table. And the next predicted value is registered, and at the time of the next observation, the change amount up to the next time is calculated and registered in the record of the change rate stage.

FIG. 13 shows a change amount sum data management table on the auxiliary storage device of the computer having the change amount stage, the current measured value, the change amount up to the next time, the next predicted value and the prediction accuracy as items. . The meaning of each item is the same as that of the Y area history data management table of FIG.
At the time of this observation, the sum of the amounts of change in the measured values observed this time from the start of measurement is calculated, and the amount of change is recorded in the record of the amount of change to which the total of the amounts of change calculated in the total amount of change data management table belongs. The number, the observed measurement value, and the next predicted value are registered, and at the time of the next observation, the sum total of the change amounts up to the next time is calculated and registered in the record of the change amount stage.

The change amount / prediction data storage means 14 includes a Y area history data management table (FIG. 10), an X area history data management table (FIG. 11), a change rate data management table (FIG. 12), and a change amount total management table. The data registered in (FIG. 13) is accumulated in the storage device 16.

The prediction accuracy calculation means 15 evaluates the difference between the predicted value and the actually observed measurement value to calculate the prediction accuracy.
That is, the Y area history data management table (FIG. 10), the X area history data management table (FIG. 11), and the change rate data management table (FIG. 12) accumulated in the storage device 16.
And in the change amount sum management table (FIG. 13), based on the current measured value, the change amount up to the next time, and the next predicted value,
The following calculation is performed and the calculation result is registered in the corresponding record of each table as the prediction accuracy.

When "current measured value + change amount up to next" and "next predicted value" are not equal: prediction accuracy = 1 / (| next predicted value- (current measured value + change amount up to next) ) |) "Current measurement value + change amount up to next" and "next predicted value"
When is equal to: Prediction accuracy = 1 Here, an example of calculation of the change amount and prediction accuracy up to the next registered in the Y region history data management table of FIG. 10 will be described. In the Y area having the Y area number 2, if the current measured value is 31, the next measured value is 43, and the next predicted value is 46, the change amount up to the next is 43.
Since −31 = 12, which corresponds to the above case,
The prediction accuracy is 1 / (46- (31 + 12)) = 0.33.

The change amount / prediction accuracy detecting means 17 performs the following four processes at the time of observing a certain measured value. The order of performing this process is not particularly limited.

The Y area history data management table having the same Y area number as the Y area to which the observed measurement value belongs (see FIG. 1).
The record group of 0) is classified for each record having the same change amount value up to the next, the data having the largest number of records is searched from the classified data, and the record in the data having the largest number of records is searched. Extract the amount of change and the prediction accuracy. In this case, if there are a plurality of data items having the largest number of records, the data item having the larger average value of the prediction accuracy is selected. Also, if the corresponding record does not exist,
The record closest to the record is extracted.

The X area history data management table having the same X area number as that of the X area to which the observed measurement value belongs (see FIG. 1).
The record group of 1) is classified for each record having the same change amount value up to the next, the data with the largest number of records is searched from the classified data, and the record change in the data with the largest number of records is searched. Extract quantity and prediction accuracy. In this case, if there are a plurality of data items having the largest number of records, the data item having the larger average value of the prediction accuracy is selected. If the corresponding record does not exist, the record closest to the record is extracted.

The record group of the change rate data management table (FIG. 12), which is the same as the change rate stage of the change rate to which the observed measurement value belongs, is classified for each record having the same change value up to the next, and classified. The data having the largest number of records is searched from the obtained data, and the change amount of the record and the prediction accuracy in the data having the largest number of records are extracted. In this case, if there are a plurality of data items having the largest number of records, the data item having the larger average value of the prediction accuracy is selected.
If the corresponding record does not exist, the record closest to the record is extracted.

The record group of the change amount total data management table (FIG. 13) which is the same as the change amount stage of the sum total of the change amounts to which the observed measurement value belongs is classified into records having the same change amount values up to the next. The data having the largest number of records is searched from the classified data, and the amount of change and the prediction accuracy of the records in the data having the largest number of records are extracted. In this case, if there are a plurality of data items having the largest number of records, the data item having the larger average value of the prediction accuracy is selected. If the corresponding record does not exist, the record closest to the record is extracted.

The change amount / prediction accuracy detecting means 14 may perform the following four processes instead of the above four processes.

The Y area history data management table (FIG. 1) having the same Y area number as the Y area to which the observed measurement value belongs
The record group of 0) is extracted, and the average value of the change amount to the next and the average value of the prediction accuracy in the extracted record group are calculated.

The X area history data management table having the same X area number as that of the X area to which the observed measurement value belongs (see FIG. 1).
The record group of 1) is extracted, and the average value of the change amount to the next and the average value of the prediction accuracy in the extracted record group are calculated.

The record group of the change rate data management table (FIG. 12) which is the same as the change rate stage of the change rate to which the observed measurement value belongs is extracted, and the average value of the change amount up to the next in the extracted record group and Calculate the average value of the prediction accuracy.

The record group of the total change amount data management table (FIG. 13) that is the same as the total change amount stage of the total amount of change to which the observed measurement value belongs is extracted, and the change amount up to the next in the extracted record group is extracted. Calculate the average value and the average value of the prediction accuracy.

The change amount calculation means 18 performs the following calculation on the four sets of change amounts and prediction accuracy detected by the change amount / prediction accuracy detection means 17 to calculate the change amount up to the next observation time. .

Change amount = Change amount 1 × Prediction accuracy 1 / (Prediction accuracy 1 + 2 + 3 + 4) + Change amount 2 × Prediction accuracy 2 / (Prediction accuracy 1 + 2 + 3 + 4) + Change amount 3 × Prediction accuracy 3 / (Prediction accuracy 1 + 2 + 3 + 4) + Change amount 4 × prediction accuracy 4 / (prediction accuracy 1 + 2 + 3 + 4) Here, the change amount 1 and the prediction accuracy 1 are detected from the Y area history data management table (FIG. 10), and the change amount 2 and the prediction accuracy 2 are the X area history data management table. The change amount 3 and the prediction accuracy 3 are detected from the change rate data management table (FIG. 12), and the change amount 4 and the prediction accuracy 4 are detected from the change amount total data management table (FIG. 13). .

Instead of performing the above calculation, the change amount calculating means 18 determines the change amount having the greatest prediction accuracy among the four prediction accuracy detected by the change amount / prediction accuracy detecting means 18 at the next observation time. It may be a change amount up to.

The data predicting means 19 is the variation calculating means 1
The next predicted value is obtained by adding the measured value to the amount of change calculated by 8.

As described above, according to the first embodiment,
Since the amount of change in the past data is weighted by the past prediction accuracy to calculate the predicted value based on the region and the stage to which the observed measurement value belongs, accurate data prediction can be performed.

In the first embodiment, the auxiliary storage device is used as the storage means of each table, but instead of this, a main storage device may be used, or a list structure may be used instead of the table. The storage means used may be used.

(Second Embodiment) A data predicting apparatus according to the second embodiment of the present invention will be described below with reference to FIGS.
It will be described based on. The second embodiment is also a data prediction device that uses a computer to predict fluctuations in the number of particles generated from a semiconductor manufacturing device. FIG. 10 is the same as that used in the first embodiment.

FIG. 2 shows a system block diagram of the data fluctuation prediction apparatus according to the second embodiment.

The data variation predicting apparatus according to the second embodiment, as shown in FIG.
The prediction data storage means 24, the prediction accuracy calculation means 25, the storage device 26, the change amount / prediction accuracy detection means 27, the change amount calculation means 28, and the data prediction means 29 predict future fluctuations of particle data observed in time series. It is a device that does. Hereinafter, each means will be described in detail.

The Y area determining means 20 is similar to the Y area determining means 10 in the first embodiment.

The change / prediction data storage means 24 has a first
Similar to the embodiment of the above, at the time of this observation, the Y region number, the current measured value, and the next predicted value are recorded in the record of the Y region to which the measured value observed this time in the Y region history data management table (FIG. 10) belongs. At the same time as the registration, at the time of the next observation, the change amount up to the next time is calculated and registered in the record of the Y area, and the data registered in the Y area history data management table (FIG. 10) is accumulated in the storage device 26. .

The prediction accuracy calculation means 25 evaluates the difference between the predicted value and the actually measured value, and calculates the prediction accuracy.
That is, the following calculation is performed based on the current measurement value, the change amount up to the next time, and the next predicted value in the Y area history data management table (FIG. 10) accumulated in the storage device 16, and the calculation result is obtained. The prediction accuracy is registered in the corresponding record in the Y area history data management table (FIG. 10).

When “current measurement value + change amount up to next time” and “next prediction value” are not equal to each other: prediction accuracy = 1 / (| next prediction value− (current measurement value + change amount up to next time) ) |) "Current measurement value + change amount up to next" and "next predicted value"
When is equal to: Prediction accuracy = 1 The change amount / prediction accuracy detecting means 27 carries out the following processing when a certain measurement value is observed.

The Y area history data management table (FIG. 1) having the same Y area number as the Y area to which the observed measurement value belongs
The record group of 0) is classified for each record having the same change amount value up to the next, the data having the largest number of records is searched from the classified data, and the record in the data having the largest number of records is searched. Extract the amount of change and the prediction accuracy. In this case, if there are a plurality of data with the largest number of records, the data with the larger average value of the prediction accuracy is selected. Also, if the corresponding record does not exist,
The record closest to the record is extracted.

The change amount calculation means 28 calculates the change amount and the prediction accuracy by the change amount / prediction accuracy detection means 27.
The amount of change until the next observation time is calculated by performing the following calculation. Change amount = Change amount 1 x Prediction accuracy 1 / (Prediction accuracy 1 + 2 + ... + n) + Change amount 2 x Prediction accuracy 2 / (Prediction accuracy 1 + 2 + ... + n) + ... + Change amount n x Prediction accuracy n / (Prediction accuracy 1 + 2 + ... + n) Here, n is the number of data detected by the variation / prediction accuracy detection means 27.

The data predicting means 29 is the variation calculating means 2
The next predicted value is obtained by adding the measured value to the amount of change calculated by 8.

(Third Embodiment) A data predicting apparatus according to a third embodiment of the present invention will be described below with reference to FIGS.
It will be described based on. The third embodiment is also a data predicting apparatus that predicts fluctuations in the number of particles generated from a semiconductor manufacturing apparatus using a computer. About FIG. 11, it is the same as that used in the first embodiment.

FIG. 3 shows a system block diagram of a data fluctuation prediction apparatus according to the third embodiment.

The data fluctuation predicting apparatus according to the third embodiment, as shown in FIG.
The prediction data accumulating unit 34, the prediction accuracy calculating unit 35, the storage device 36, the change amount / prediction accuracy detecting unit 37, the change amount calculating unit 38, and the data predicting unit 39 predict future fluctuations of particle data observed in time series. It is a device that does.

The X area determining means 31 is the same as the X area determining means 11 in the first embodiment.

Further, the change / prediction data storage means 34,
The prediction accuracy calculation unit 35, the storage device 36, the change amount / prediction accuracy detection unit 37, the change amount calculation unit 38, and the data prediction unit 39 are the Y area and the Y area history data management table (FIG. 10) in the second embodiment. Instead of using the X area and the X area history data management table (FIG. 11), the description is omitted because it is the same as the second embodiment.

(Fourth Embodiment) Hereinafter, a data prediction apparatus according to a fourth embodiment of the present invention will be described with reference to FIGS.
It will be described based on. The fourth embodiment is also a data predicting apparatus that predicts fluctuations in the number of particles generated from a semiconductor manufacturing apparatus using a computer. FIG. 12 is the same as that used in the first embodiment.

FIG. 4 shows a system block diagram of a data variation predicting apparatus according to the fourth embodiment.

The data fluctuation predicting apparatus according to the fourth embodiment, as shown in FIG. 4, includes a change rate step dividing means 42, a change amount / predicted data accumulating means 44, a predictive accuracy calculating means 45,
The storage device 46, the change amount / prediction accuracy detecting means 47, the change amount calculating means 48, and the data predicting means 49 are devices for predicting future fluctuations of particle data observed in time series.

The change rate step dividing means 42 is the same as the change rate step dividing means 12 in the first embodiment.

The change / prediction data storage means 44,
The prediction accuracy calculation unit 45, the storage device 46, the change amount / prediction accuracy detection unit 47, the change amount calculation unit 48, and the data prediction unit 49 are the Y area and Y area history data management table (FIG. 10) in the second embodiment. Instead of the change rate stage and the change rate management table (FIG. 12), the second embodiment is the same as the second embodiment, and therefore the description thereof is omitted.

(Fifth Embodiment) Hereinafter, a data prediction apparatus according to a fifth embodiment of the present invention will be described with reference to FIGS.
It will be described based on. The fifth embodiment is also a data predicting apparatus that predicts fluctuations in the number of particles generated from a semiconductor manufacturing apparatus using a computer. 13 is the same as that used in the first embodiment.

FIG. 5 shows a system block diagram of the data variation prediction apparatus according to the fifth embodiment.

The data fluctuation predicting apparatus according to the fifth embodiment, as shown in FIG. 5, has a change amount stage dividing means 53, a change amount / predicted data accumulating means 54, a prediction accuracy calculating means 55,
The storage device 56, the change amount / prediction accuracy detection unit 57, the change amount calculation unit 58, and the data prediction unit 59 are devices for predicting future fluctuations of particle data observed in time series.

The change amount stepping means 53 is the same as the change amount stepping means 13 in the first embodiment.

Further, the change amount / prediction data storage means 54,
The prediction accuracy calculation unit 55, the storage device 56, the change amount / prediction accuracy detection unit 57, the change amount calculation unit 58, and the data prediction unit 59 are the Y area and Y area history data management table (FIG. 10) in the second embodiment. Instead of using the change amount stage and the change amount total management table (FIG. 13), the description is omitted because it is the same as the second embodiment.

As described above, according to the second to fifth embodiments, based on the region or step to which the observed measurement value belongs,
Since the predicted value is calculated by weighting the change amount of the past data with the past prediction accuracy, it is possible to accurately predict the data.

In the second to fifth embodiments, the auxiliary storage device is used as the storage means of each table, but instead of this, a main storage device may be used, or instead of the table. A storage means using a list structure may be used.

(Sixth Embodiment) A data predicting apparatus according to a sixth embodiment of the present invention will be described below with reference to FIGS.
This will be described with reference to FIG. The sixth embodiment is also a data predicting apparatus for predicting fluctuations in the number of particles generated from a semiconductor manufacturing apparatus using a computer. Figure 10
13 is the same as that used in the first embodiment.

FIG. 6 is a block diagram showing the system of the data variation prediction apparatus according to the sixth embodiment. The data variation prediction apparatus according to the sixth embodiment, as shown in FIG.
Y area determining means 60, X area determining means 61, change rate step dividing means 62, change amount step dividing means 63, change amount accumulating means 64, storage device 65, change amount detecting means 66, change amount calculating means 67 and data prediction. The means 68 is a device for predicting future changes in particle data that are managed in the XY coordinate system and are observed in time series. Hereinafter, each means will be described in detail.

Y area determining means 60, X area determining means 6
1. Change rate stepping means 62 and change amount stepping means 6
3 is the same as the Y area determining means 10, the X area determining means 11, the change rate step dividing means 12 and the change amount step dividing means 13 in the first embodiment.

The change amount accumulating means 64 carries out the following four processes when a certain measured value is observed. The order of performing this process is not particularly limited.

At the time of this observation, the Y area number and the current measurement value are registered in the record of the Y area to which the measurement value observed this time in the Y area history data management table (FIG. 10) is registered, and at the time of the next observation time. In, the amount of change until the next time is calculated and registered in the record of the Y area.

At the time of this observation, the X area number and the current measurement value are registered in the record of the X area to which the measurement value observed this time in the X area history data management table (FIG. 11) is registered. At the time of observation, the change amount up to the next time is calculated and registered in the record of the X area.

At the time of this observation, the change rate stage and the current measured value are registered in the record of the change rate stage to which the measured value observed this time belongs in the change rate history data management table (FIG. 12). At the time of observation, the amount of change until the next time is calculated and registered in the record of the change rate stage.

At the time of this observation, the change amount stage and the current measurement value are registered in the record of the change amount stage to which the measurement value observed this time belongs in the change amount total history data management table (FIG. 13). At the time of the next observation, the change amount up to the next time is calculated and registered in the record of the change amount stage.

The change amount accumulating means 64 has a Y area history data management table (FIG. 10), an X area history data management table (FIG. 11), a change rate data management table (FIG. 12), and a change amount total management table (FIG. 13). The data registered in) are accumulated in the storage device 65.

The change amount detecting means 66 carries out the following four processes when a certain measured value is observed. The order of performing this process is not particularly limited.

The Y area history data management table having the same Y area number as the Y area to which the observed measurement value belongs (see FIG. 1).
The record group of 0) is classified for each record having the same change amount value up to the next, the data having the largest number of records is searched from the classified data, and the record in the data having the largest number of records is searched. Extract the amount of change. In this case, if there are a plurality of data items having the largest number of records, the data item having the larger change amount is selected. If the corresponding record does not exist, the record closest to the record is extracted.

The X area history data management table having the same X area number as that of the X area to which the observed measurement value belongs (see FIG. 1).
The record group of 1) is classified for each record having the same change amount value up to the next, the data with the largest number of records is searched from the classified data, and the record change in the data with the largest number of records is searched. Extract the amount. In this case, if there are a plurality of data items having the largest number of records, the data item having the larger change amount is selected. If the corresponding record does not exist, the record closest to the record is extracted.

The record group of the change rate data management table (FIG. 12) that is the same as the change rate stage of the change rate to which the observed measurement value belongs is classified for each record having the same change amount value up to the next, and classified. The data with the largest number of records is searched from the obtained data, and the change amount of the record in the data with the largest number of records is extracted. In this case, if there are a plurality of data items having the largest number of records, the data item having the larger change amount is selected. If the corresponding record does not exist, the record closest to the record is extracted.

The record group of the total change amount data management table (FIG. 13), which is the same as the total change amount stage of the total change amounts to which the observed measurement value belongs, is classified into records having the same change amount values up to the next. Then, the data having the largest number of records is searched from the classified data, and the change amount of the record in the data having the largest number of records is extracted. In this case, if there are a plurality of data items having the largest number of records, the data item having the larger change amount is selected. If the corresponding record does not exist, the record closest to the record is extracted.

The change amount detecting means 66 may perform the following four processes instead of the above four processes.

The Y area history data management table having the same Y area number as the Y area to which the observed measurement value belongs (see FIG. 1).
The record group of 0) is extracted, and the average value of the change amounts up to the next in the extracted record group is calculated.

The X area history data management table having the same X area number as that of the X area to which the observed measurement value belongs (see FIG.
The record group of 1) is extracted, and the average value of the change amounts up to the next time in the extracted record group is calculated.

The record group of the change rate data management table (FIG. 12) that is the same as the change rate stage of the change rate to which the observed measurement value belongs is extracted, and the average value of the change amounts up to the next in the extracted record group is calculated. calculate.

The record group of the total change amount data management table (FIG. 13) which is the same as the total change amount stage of the total amounts of change to which the observed measured value belongs is extracted, and the change amount up to the next in the extracted record group is extracted. Calculate the average value.

The change amount calculating means 67 is the change amount detecting means 6
The average of the four changes detected by 6 is calculated, and the change until the next observation point is calculated.

The data predicting means 68 is the variation calculating means 6
The next predicted value is obtained by adding the measured value to the variation calculated by 7.

As described above, according to the sixth embodiment, four fluctuation factors (X area, Y area, rate of change, total change amount) are added based on the area and stage to which the observed measurement value belongs. Since the predicted value is calculated by the above, accurate data prediction can be performed.

In the sixth embodiment, the auxiliary storage device is used as the storage means of each table, but instead of this, a main storage device may be used, or a list structure may be used instead of the table. The storage means used may be used.

(Seventh Embodiment) A data prediction method according to the seventh embodiment of the present invention will be described below with reference to FIGS. 7, 9 and 1.
4 and FIG. 15. The seventh embodiment is
This is a method of predicting fluctuations in the number of particles generated from a semiconductor manufacturing apparatus.

FIG. 7 shows a processing flow of the data variation prediction method according to the seventh embodiment, FIG. 9 shows an example of device particle data managed in the XY coordinate system, and FIG.
Shows the processing of the change amount / prediction data storage step (step A5) in the data variation prediction method according to the seventh embodiment, and FIG. 15 is the change amount / prediction accuracy in the data variation prediction method according to the seventh embodiment. Detection step (Step A
The process of 7) is shown. Hereinafter, each step will be described in detail.

In step A1, the range from the minimum value to the upper limit value of the number of particles detected by the semiconductor processing apparatus is set as the Y region, and the Y region is set to, for example, 6
Y, which is a number for identifying each of the divided and divided areas
Assign a region number. Note that the number of divisions of the Y region is not limited, and the Y region may be divided into an appropriate number in consideration of the maximum value of the number of particles, the prediction accuracy, and the like.

In step A2, the range in which the number of processed wafers detected by the semiconductor processing apparatus can be taken in the X-axis direction is defined as an X area, and the X area is divided at predetermined intervals, for example, in units of 100 wafers. An X area number, which is a number for identification, is given to each of the created areas. The unit for dividing the X region is not limited, and the dividing unit can be determined in consideration of the maximum value of the number of processed wafers in the past, the particle count measurement frequency, the prediction accuracy, and the like.

In step A3, the rate of change of the number of particles with respect to the number of processed wafers detected by the semiconductor processing apparatus is divided into predetermined steps, for example, 16 steps, and each divided step is a number for identification. A rate of change step number is assigned. The number of divisions of the change rate is not limited, and the number of divisions can be determined in consideration of the number of processed wafers, the particle number measurement frequency, the prediction accuracy, and the like.

In step A4, the total amount of change in the number of particles with respect to the number of processed wafers detected by the semiconductor process apparatus is divided into predetermined steps, for example, 60 steps, and a number for identification is given to each of the divided steps. The change amount step number is given. The number of divisions of the total sum of change amounts is not limited, and the number of divisions can be determined in consideration of the maximum value of the total sum of past particle numbers, the particle number measurement frequency, the prediction accuracy, and the like. Further, the sum of change amounts may be the sum of absolute values of the respective change amounts, or the sum of relative values in consideration of + -of the change amounts.

The order of performing step A1, step A2, step A3 and step A4 is not particularly limited.

Next, at step A5, when a certain measured value is observed, four processes shown in FIG. 14 are performed.
The order of performing this process is not particularly limited.

The observed measurement value is stored for each Y region together with the predicted value at the next observation time point. Further, the amount of change obtained by subtracting the measurement value of the previous observation from the measurement value of the actual observation at the time of the next observation is calculated and stored.

The observed measurement value is stored for each X region together with the predicted value at the next observation time point. Further, the amount of change obtained by subtracting the measurement value of the previous observation from the measurement value of the actual observation at the time of the next observation is calculated and stored.

The observed measurement value is stored for each change rate step together with the predicted value at the next observation time point. Further, the amount of change obtained by subtracting the measurement value of the previous observation from the measurement value of the actual observation at the time of the next observation is calculated and stored.

The observed measurement value is stored together with the predicted value at the next observation point for each change amount step. Further, the amount of change obtained by subtracting the measurement value of the previous observation from the measurement value of the actual observation at the time of the next observation is calculated and stored.

As the predicted value at the next observation time, the predicted value predicted in step A9 described later is used.

Next, in step A6, the difference between the predicted value and the actually observed value is evaluated to calculate the prediction accuracy. That is, the following calculation is performed based on the measured value, the change amount, and the predicted value stored in step A5,
The calculation result is stored as the prediction accuracy.

When “measured value + change amount” and “predicted value” are not equal: prediction accuracy = 1 / (| predicted value− (measured value + change amount) |) “measured value + change amount” and “predicted value” Prediction accuracy = 1 Next, at step A7, when a certain measured value is observed, four processes shown in FIG. 15 are performed. The order of performing this process is not particularly limited.

From the data stored in the Y area to which the observed measurement values belong, the one with the largest number of data with the same change amount value is searched for, and the change amount and prediction accuracy are extracted.
If there are a plurality of data with the same maximum number of data, the data with the larger average value of the prediction accuracy is selected. If the corresponding data does not exist, the data is extracted from the Y area in which the data closest to the data exists.

From the data stored in the X area to which the observed measurement values belong, the one with the largest number of data having the same change amount value is searched for, and the change amount and prediction accuracy are extracted.
If there are a plurality of data with the same maximum number of data, the data with the larger average value of the prediction accuracy is selected. If the corresponding data does not exist, the data is extracted from the X area in which the data closest to the data exists.

From the data stored in the change rate stage to which the observed measurement value belongs, the one having the largest number of data having the same change amount value is searched for, and the change amount and the prediction accuracy are extracted. If there are multiple data with the same maximum number of data,
Select the data with the larger average value of the prediction accuracy. If there is no corresponding data, the data is extracted from the change rate stage where the data closest to the data exists.

From the data stored at the change amount stage to which the observed measurement value belongs, the one having the largest number of data having the same change amount value is searched for, and the change amount and the prediction accuracy are extracted. If there are multiple data with the same maximum number of data,
Select the data with the larger average value of the prediction accuracy. If there is no corresponding data, the data is extracted from the change amount stage in which the data closest to the data exists.

In step A7, the following four processes may be performed instead of the above four processes.

The average value of the amount of change and the average value of the prediction accuracy of the data stored in the Y region to which the observed measurement value belongs are calculated.

The average value of the amount of change of the data stored in the X region to which the observed measurement value belongs and the average value of the prediction accuracy are calculated.

The average value of the amount of change and the average value of the prediction accuracy of the data stored in the change rate stage to which the observed measurement value belongs are calculated.

The average value of the change amounts of the data stored in the change amount stage to which the observed measurement value belongs and the average value of the prediction accuracy are calculated.

In step A8, the following calculation is performed on the four sets of change amounts and prediction accuracy extracted in step A7 to calculate the change amount up to the next observation time point.

Change amount = Change amount 1 × Prediction accuracy 1 / (Prediction accuracy 1 + 2 + 3 + 4) + Change amount 2 × Prediction accuracy 2 / (Prediction accuracy 1 + 2 + 3 + 4) + Change amount 3 × Prediction accuracy 3 / (Prediction accuracy 1 + 2 + 3 + 4) + Change amount 4 × prediction accuracy 4 / (prediction accuracy 1 + 2 + 3 + 4) Here, the variation 1 and the prediction accuracy 1 are detected from the Y region,
The change amount 2 and the prediction accuracy 2 are detected from the X region, and the change amount 3
And the prediction accuracy 3 is detected from the change rate stage, and the change amount 4 and the prediction accuracy 4 are detected from the change amount stage.

In step A8, the amount of change having the greatest prediction accuracy among the four prediction accuracy extracted in step A7 may be used as the amount of change until the next observation time, instead of the above calculation.

At step A9, by adding the measured value to the variation calculated at step A8,
Find the next predicted value.

As described above, according to the seventh embodiment,
Since the amount of change in the past data is weighted by the past prediction accuracy to calculate the predicted value based on the region and the stage to which the observed measurement value belongs, accurate data prediction can be performed.

(Eighth Embodiment) Hereinafter, a data prediction method according to an eighth embodiment of the present invention will be described with reference to FIGS.
6 and FIG. 17. The eighth embodiment is
This is a method of predicting fluctuations in the number of particles generated from a semiconductor manufacturing apparatus.

FIG. 8 shows the flow of processing of the data variation prediction method according to the eighth embodiment of the present invention, and FIG. 16 shows the storage process (step) of the variation in the data variation prediction method according to the eighth embodiment. B5) is shown, and FIG. 17 shows a change amount detecting step (step B6) in the data variation prediction method according to the eighth embodiment.

Step B1, Step B2, Step B
3 and step B4 are the same as step A1, step A2, step A3 and step A4 in the seventh embodiment. The order of performing Step B1, Step B2, Step B3, and Step B4 is not particularly limited.

At step B5, when a certain measured value is observed, the four processes shown in FIG. 16 are performed. The order of performing this process is not particularly limited.

The observed measurement value is stored for each Y region, and the change amount obtained by subtracting the measurement value of the previous observation from the measurement value of the actual observation at the next observation time is stored.

The observed measurement value is stored for each X region, and the change amount obtained by subtracting the measurement value at the previous observation from the measurement value actually observed at the next observation time is stored.

The observed measurement value is stored for each change rate step, and the change amount obtained by subtracting the measurement value at the previous observation from the measurement value actually observed at the next observation time is stored.

The observed measurement value is stored for each change amount step, and the change amount obtained by subtracting the measurement value at the previous observation from the measurement value actually observed at the next observation time is stored.

At step B6, when a certain measured value is observed, the four processes shown in FIG. 17 are performed. The order of performing this process is not particularly limited.

From the data stored in the Y area to which the observed measurement value belongs, the one having the largest number of data pieces having the same change amount value is searched and the change amount is extracted. If there are a plurality of pieces of data having the same maximum number of pieces of data, the data having the larger amount of change is selected. If the corresponding data does not exist, the data is extracted from the Y area in which the data closest to the data exists.

From the data stored in the X region to which the observed measurement value belongs, the one having the largest number of data pieces having the same change amount value is searched and the change amount is extracted. If there are a plurality of pieces of data having the same maximum number of pieces of data, the data having the larger amount of change is selected. If the corresponding data does not exist, the data is extracted from the X area in which the data closest to the data exists.

From the data stored in the change rate stage to which the observed measurement value belongs, the one having the largest number of data pieces having the same change amount value is searched and the change amount is extracted. If there are a plurality of pieces of data having the same maximum number of pieces of data, the data having the larger amount of change is selected. If there is no corresponding data, the data is extracted from the change rate stage where the data closest to the data exists.

From the data stored in the change amount stage to which the observed measurement value belongs, the one having the largest number of data items having the same change amount value is searched and the change amount is extracted. If there are a plurality of pieces of data having the same maximum number of pieces of data, the data having the larger amount of change is selected. If there is no corresponding data, the data is extracted from the change amount stage in which the data closest to the data exists.

In step B6, the following four processes may be performed instead of the above four processes.

The average value of the amount of change in the data stored in the Y region to which the observed measurement value belongs is calculated.

The average value of the amount of change in the data stored in the X region to which the observed measurement value belongs is calculated.

The average value of the amount of change of the data stored in the change rate stage to which the observed measurement value belongs is calculated.

The average value of the change amounts of the data stored in the change amount stage to which the observed measurement value belongs is calculated.

In step B7, the average value of the four sets of change amounts extracted in step B6 is calculated, and the change amount up to the next observation point is calculated.

In step B8, the next predicted value is obtained by adding the measured value observed to the amount of change calculated in step B7.

As described above, according to the eighth embodiment,
Based on the region and the stage to which the observed measurement value belongs, the prediction value is calculated by adding the four fluctuation factors (X region, Y region, change rate, and total change amount), so that accurate data prediction is performed. be able to.

[0169]

According to the data variation predicting apparatus of the first aspect of the present invention, the predicted value is obtained by weighting the variation amount of each measured value divided area with the prediction accuracy of each measured value divided area. Data can be created in a short time without providing historical data.

According to the data variation prediction apparatus of the second aspect of the present invention, since the variation amount for each observation number division region is weighted by the prediction accuracy for each observation number division region to obtain a prediction value, accurate data is obtained. Can be created in a short time without providing history data.

According to the data fluctuation predicting apparatus of the third aspect of the present invention, accurate data is stored in the history because the predicted value is obtained by weighting the change amount at each change rate step with the prediction accuracy at each change rate step. It can be created in a short time without data.

According to the data fluctuation predicting apparatus of the fourth aspect of the present invention, since the predicted value is obtained by weighting the change amount of each change amount stage with the prediction accuracy of each change amount stage, accurate data is recorded in the history. It can be created in a short time without data.

According to the data variation prediction apparatus of the fifth aspect of the present invention, the measurement value division area, the observation number division area, the change rate step and the change amount for each change step are divided into the measurement value division area and the observation number division. Since the predicted value is obtained by weighting the prediction accuracy for each region, change rate step, and change amount step, more accurate data can be created in a short time without providing historical data.

According to the data fluctuation predicting apparatus of the sixth aspect of the present invention, since the prediction value created by the prediction value creating means is used as the prediction value at the time of the second measurement in the prediction accuracy creating means, the accuracy of the prediction accuracy is improved. To do.

According to the data fluctuation predicting apparatus of the seventh aspect of the present invention, the average value of change amounts belonging to the measurement value division region, the observation number division region, the change rate stage and the change amount stage to which the measurement value observed this time belongs On the other hand, since the weighting is performed by the average value of the prediction accuracy belonging to the measurement value division area to which the measurement value observed this time belongs, the observation number division area, the change rate step and the change amount step, the accuracy of the obtained prediction value is further improved. To do.

According to the data variation prediction apparatus of the eighth aspect of the invention, the prediction accuracy = 1 / | the prediction at the second measurement for each of the measurement value division area, the observation number division area, the change rate step and the change amount step. Since the prediction accuracy is created by the formula value-second measurement value |, the prediction accuracy decreases when the difference between the prediction value and the measurement value is large, and the prediction accuracy increases when the difference between the prediction value and the measurement value is small. Therefore, the difference between the predicted value and the measured value is accurately reflected in the prediction accuracy.

According to the data fluctuation predicting apparatus of the ninth aspect of the present invention, the prediction accuracy with high accuracy contributes greatly to the calculation of the amount of change, while the prediction accuracy with low accuracy contributes little. Since the quantity is reliably reflected, the accuracy of the predicted value is improved.

According to the data fluctuation predicting apparatus of the tenth aspect of the present invention, the measurement value division area, the observation number division area among the change amounts of the measurement value division area, the observation number division area, the change rate step and the change amount step. , The prediction accuracy corresponding to the largest prediction accuracy of the change rate step and the change amount step is selected and calculated, so the change quantity corresponding to the largest prediction accuracy is calculated. The accuracy of the value is improved.

According to the data variation prediction apparatus of the eleventh aspect of the present invention, the change frequently belonging to the measurement value division area, the observation number division area, the change rate stage and the change amount stage to which the measurement value observed this time belongs respectively. In order to create the predicted value for the next observation by adding the average change amount, which is the average value of the amount, to the measured value observed this time, the predicted value obtained is the measured value division area, the observation number division area, the change rate stage, Since the calculation is performed based on the four variation factors in the change amount stage, more accurate data can be created in a short time without providing historical data.

According to the data fluctuation prediction apparatus of the twelfth aspect of the present invention, the average value of each change amount belonging to the measurement value division region, the observation number division region, the change rate stage and the change amount stage to which the measurement value observed this time belongs. Is added to the measured value observed this time to create the predicted value for the next observation, the predicted value obtained is the measured value division area, the observation number division area, and the rate of change step. Since it is calculated on the basis of four fluctuation factors of the change amount stage and the change amount stage, accurate data can be created in a short time without providing historical data.

According to the data fluctuation prediction method of the thirteenth aspect, each change that belongs to the measurement value division region, the observation number division region, the change rate stage and the change amount stage, and has a high frequency, belongs to the measurement value observed this time. The amount of change created by weighting the amount with prediction accuracy is added to the measured value observed this time to create the predicted value for the next observation. Since it is calculated based on the four variable factors of the divided region, the change rate stage and the change amount stage and is weighted by the prediction accuracy, more accurate data can be created without providing historical data. .

According to the data fluctuation prediction method of the fourteenth aspect of the present invention, the average value of each change amount belonging to the measurement value division region, the observation number division region, the change rate stage and the change amount stage to which the measurement value observed this time belongs. Since the amount of change created by weighting with respect to the prediction accuracy is added to the measured value observed this time to create the predicted value for the next observation, the obtained predicted value is divided into the measurement value division area and the observation number division. Since it is calculated based on four variable factors of the region, the rate of change stage, and the amount of change stage, and weighted by the prediction accuracy, more accurate data can be created without providing historical data.

According to the data variation prediction method of the fifteenth aspect of the present invention, the prediction accuracy = 1 / | the prediction at the second measurement for each of the measurement value division area, the observation number division area, the change rate step and the change amount step. Since the prediction accuracy is created by the formula value-second measurement value |, the prediction accuracy decreases when the difference between the prediction value and the measurement value is large, and the prediction accuracy increases when the difference between the prediction value and the measurement value is small. Therefore, the difference between the predicted value and the measured value is accurately reflected in the prediction accuracy.

According to the data fluctuation prediction method of the sixteenth aspect of the present invention, the prediction accuracy with high accuracy contributes greatly to the calculation of the variation amount, while the prediction accuracy with low accuracy contributes little. Since the quantity is reliably reflected, the accuracy of the predicted value is improved.

According to the data variation prediction method of the seventeenth aspect of the present invention, the measurement value division region and the observation number division region are selected from the measurement value division region, the observation number division region, and the change amounts of the change rate stage and the change amount stage. , The prediction accuracy corresponding to the largest prediction accuracy of the change rate step and the change amount step is selected and calculated, so the change quantity corresponding to the largest prediction accuracy is calculated. The accuracy of the value is improved.

According to the data variation prediction method of the eighteenth aspect of the present invention, a change frequently belonging to each of the measurement value division area, the observation number division area, the change rate stage and the change amount stage to which the measurement value observed this time belongs In order to create the predicted value for the next observation by adding the average change amount, which is the average value of the amount, to the measured value observed this time, the predicted value obtained is the measured value division area, the observation number division area, the change rate stage, Since the calculation is performed based on the four variation factors in the change amount stage, more accurate data can be created in a short time without providing historical data.

According to the data fluctuation prediction method of the nineteenth aspect of the present invention, the average value of each change amount belonging to the measurement value division region, the observation number division region, the change rate stage and the change amount stage to which the measurement value observed this time belongs. Is added to the measured value observed this time to create the predicted value for the next observation, the predicted value obtained is the measured value division area, the observation number division area, and the rate of change step. Since it is calculated on the basis of four fluctuation factors of the change amount stage and the change amount stage, accurate data can be created in a short time without providing historical data.

[Brief description of drawings]

FIG. 1 is a system block diagram of a data variation prediction device according to a first embodiment of the present invention.

FIG. 2 is a system block diagram of a data variation prediction device according to a second embodiment of the present invention.

FIG. 3 is a system block diagram of a data variation prediction device according to a third embodiment of the present invention.

FIG. 4 is a system block diagram of a data variation prediction device according to a fourth embodiment of the present invention.

FIG. 5 is a system block diagram of a data variation prediction device according to a fifth embodiment of the present invention.

FIG. 6 is a system block diagram of a data variation prediction device according to a sixth embodiment of the present invention.

FIG. 7 is a flowchart showing a processing flow of a data variation prediction method according to a seventh embodiment of the present invention.

FIG. 8 is a flowchart showing the process of the data variation prediction method according to the eighth embodiment of the present invention.

FIG. 9 is data showing an example of the relationship between the number of processed wafers and the number of particles used in each embodiment of the present invention.

FIG. 10 is a diagram of a Y area history data management table used in each embodiment of the present invention.

FIG. 11 is a diagram of an X area history data management table used in each embodiment of the present invention.

FIG. 12 is a diagram of a change rate history data management table used in each embodiment of the present invention.

FIG. 13 is a diagram of a change amount total history data management table used in each embodiment of the present invention.

FIG. 14 is a diagram showing the contents of processing of a change amount / prediction data storage step in the data variation prediction method according to the seventh embodiment of the present invention.

FIG. 15 is a diagram showing the contents of processing in a change amount / prediction accuracy detection step in the data variation prediction method according to the seventh embodiment of the present invention.

FIG. 16 is a diagram showing the contents of processing of a change amount storing step in the data variation prediction method according to the eighth embodiment of the present invention.

FIG. 17 is a diagram showing the contents of processing in a change amount detection step in the data variation prediction method according to the eighth embodiment of the present invention.

FIG. 18 is a block diagram of a conventional data variation prediction device.

[Explanation of symbols]

 10 Y area determination means 11 X area determination means 12 Change rate stepping means 13 Change amount stepping means 14 Change amount / prediction data storage means 15 Prediction accuracy calculation means 16 Storage device 17 Change amount / prediction accuracy detection means 18 Change amount calculation Means 19 Data predicting means 20 Y area determining means 24 Change amount / prediction data accumulating means 25 Prediction accuracy calculating means 26 Storage device 27 Change amount / prediction accuracy detecting means 28 Change amount calculating means 29 Data predicting means 31 X area determining means 34 Change Quantity / prediction data storage means 35 Prediction accuracy calculation means 36 Storage device 37 Change amount / prediction accuracy detection means 38 Change amount calculation means 39 Data prediction means 42 Change rate stepping means 44 Change amount / prediction data storage means 45 Prediction accuracy calculation means 46 storage device 47 change amount / prediction accuracy detection means 48 change amount calculation means 4 Data predicting means 53 Change amount grading means 54 Change amount / prediction data accumulating means 55 Prediction accuracy calculating means 56 Storage device 57 Change amount / prediction accuracy detecting means 58 Change amount calculating means 59 Data predicting means 60 Y area determining means 61 X area Determining means 62 Change rate stepping means 63 Change amount stepping means 64 Change amount accumulating means 65 Storage device 66 Change amount detecting means 67 Change amount calculating means 68 Data predicting means

Claims (19)

[Claims] 1. A data fluctuation prediction apparatus for predicting future fluctuations of measured values observed in a time series, wherein the range of the measured values is divided at predetermined intervals to create a measured value division area. Region creating means, a first measurement value observed during the first observation and a second measurement value observed during the second observation after the first observation, for each of the measurement value divided regions A first change amount calculating means for calculating a first change amount which is a difference between the measured value divided region and the difference between the predicted value at the second measurement and the second measured value. Prediction accuracy creating means for creating prediction accuracy based on the first change amount calculated by the first change amount calculating means and prediction accuracy created by the prediction accuracy creating means for each of the measurement value divided areas. And a data storage means for storing From the first change amount and the prediction accuracy, a change amount / prediction accuracy detection unit that detects the first change amount and the prediction accuracy that belong to the measurement value divided region to which the measurement value observed this time belongs, and the change amount. A second change amount calculation for calculating a second change amount by weighting the first change amount detected by the prediction accuracy detection unit with the prediction accuracy detected by the change amount / prediction accuracy detection unit. Means and a predicted value creating means for creating a predicted value for the next observation by adding the second variation calculated by the second variation calculating means to the measurement value observed this time. Characteristic data fluctuation prediction device. 2. A data fluctuation predicting device for predicting future fluctuations of measured values observed in a time series, wherein the number of observed objects or the number of observations appearing in the time series is divided at predetermined intervals to divide the number of observations. Divided region creating means for creating a region, and for each of the number of observed divided regions, a first measurement value observed during the first observation and an observation during the second observation after the first observation A first change amount calculating means for calculating a first change amount which is a difference from the second measured value; a predicted value at the second measurement and the second measurement for each of the observation number division regions. A prediction accuracy creating unit that creates a prediction accuracy based on a difference from the value, and a first change amount calculated by the first change amount calculating unit and the prediction accuracy creating unit for each of the observation number division regions. Data storage means for storing the created prediction accuracy, and the data storage Change amount / prediction accuracy detection for detecting the first change amount and prediction accuracy belonging to the observation number division region to which the measurement value observed this time belongs, from the first change amount and prediction accuracy stored in the column. Means for calculating the second variation amount by weighting the first variation amount detected by the variation amount / prediction precision detecting means with the prediction precision detected by the variation amount / prediction accuracy detecting means. A second change amount calculating means, and a predicted value creating means for adding the second change amount calculated by the second change amount calculating means to the measured value observed this time to create a predicted value for the next observation. A data fluctuation prediction device comprising: 3. A data fluctuation predicting device for predicting future fluctuations of measured values observed in time series, the first measured value observed at the first observation time and the first measured value observed at the first observation time. A change rate step creating means for creating a change rate step by dividing a change rate with a second measurement value observed at a later second observation into predetermined rates, and the change rate step for each of the change rate steps. A first change amount calculating means for calculating a first change amount which is a difference between the first measured value and the second measured value; and a predicted value at the time of the second measurement for each of the change rate stages. Prediction accuracy creating means for creating prediction accuracy based on the difference from the second measured value; first change amount calculated by the first change amount calculating means and the prediction for each change rate step. Data storing means for storing the prediction accuracy created by the accuracy creating means, and the data storing means From the stored first change amount and prediction accuracy, change amount / prediction accuracy detection means for detecting the first change amount and prediction accuracy belonging to the change rate stage to which the measured value observed this time belongs, A second change amount is calculated by weighting the first change amount detected by the change amount / prediction accuracy detection unit with the prediction accuracy detected by the change amount / prediction accuracy detection unit. A change amount calculation means and a predicted value creation means for creating a predicted value for the next observation by adding the second change amount calculated by the second change amount calculation means to the measured value observed this time A data fluctuation prediction device characterized in that 4. A data fluctuation predicting device for predicting future fluctuations of measured values observed in a time series, the change from a measured value observed at the start of observation to each measured value observed at every observation time. Change amount stage creating means for creating a change amount stage by dividing the total sum of amounts into predetermined amounts, and a first measurement value observed at a first observation and the first change amount stage for each change amount stage. A first change amount calculating means for calculating a first change amount which is a difference from a second measured value observed at a second observation time after the observation time; 2 prediction accuracy creating means for creating prediction accuracy based on the difference between the predicted value at the time of measurement 2 and the second measured value, and the first change amount calculating means for each of the change amount stages. A storage for storing the first variation and the prediction accuracy created by the prediction accuracy creating means. Data storage means and the first variation and prediction accuracy stored in the data storage means, the first variation and prediction accuracy belonging to the variation level to which the measurement value observed this time belongs are detected. Change amount / prediction accuracy detection means, and weighting the first change amount detected by the change amount / prediction accuracy detection means with the prediction accuracy detected by the change amount / prediction accuracy detection means. The second change amount calculation means for calculating the change amount of 2 and the second change amount calculated by the second change amount calculation means are added to the measurement value observed this time to obtain the predicted value for the next observation. A data fluctuation prediction apparatus comprising: a predicted value creation means for creating. 5. A data fluctuation predicting device for predicting future fluctuations of measured values observed in a time series, wherein a range of measured values is divided at predetermined intervals to create a measured value division area. 1 divided area creating means, a second divided area creating means for creating an observed number divided area by dividing the number of observed objects or the number of observations appearing in time series at predetermined intervals, and observed at the time of the first observation. A change in which the rate of change between the first measured value and the second measured value observed in the second observation after the first observation is divided into predetermined rates to create a change rate step. Rate step creation means and change step creation that creates a change step by dividing the total amount of change from the measurement value observed at the start of observation to each measurement value observed at each observation step by a predetermined ratio Means, the measurement value division area, the observation number division area, the change rate step and The difference between the first measurement value and the second measurement value, the measurement value division area change amount, the observation number division area change amount, the change rate step change amount, and the change amount step change amount A first change amount calculating means for calculating each of the measurement value division region, the observation number division region, the change rate step and the change amount step, and the second predicted value and the second measurement value.
Based on the difference from the measured value of
Prediction accuracy creating means for creating the number of observations divided area prediction accuracy, the rate of change step prediction accuracy, and the amount of change step prediction accuracy, respectively, the measurement value divided area, the number of observations divided area, each change rate step and each change amount step, Measurement value division area change amount calculated by the first change amount calculation means, observation number division area change amount, change rate step change amount and change amount step change amount, and measurement value division area prediction created by the prediction accuracy creation means Accuracy, the number of observations divided region prediction accuracy, the change rate step prediction accuracy, and the change amount step prediction accuracy, and a data storage unit that stores the measurement value divided region change amount and the measurement value stored in the data storage unit. From the divided area prediction accuracy, the measured value divided area change amount and the measured value divided area prediction accuracy belonging to the measured value divided area to which the measurement value observed this time belongs is detected and output, From the number of observations divided region change amount and the number of observations divided region prediction accuracy stored in the data storage means, the number of observations divided region change amount and the number of observations belonging to the number of observations divided region to which the measurement value observed this time belongs A change belonging to the change rate stage to which the measured value observed this time belongs is selected from the change rate stage change amount and the change rate stage predictive accuracy stored in the data storage means by detecting and outputting the divided area prediction precision. The rate step change amount and the change rate step prediction accuracy are detected and output, and from the change amount step change amount and the change amount step prediction accuracy stored in the data storage means, the measurement value observed this time belongs to Change amount / prediction accuracy detecting means for detecting and outputting the change amount / change amount belonging to the change amount stage and the change amount / prediction accuracy, and the measurement value division output from the change amount / prediction accuracy detecting means. Region change amount, observation number division region change amount, change rate step change amount, and change amount step change amount, the measurement value division region prediction accuracy and the observation number division region prediction output from the change amount / prediction accuracy detection means A second change amount calculating means for calculating the change amount by weighting the accuracy, the change rate step predictive accuracy and the change amount step predictive accuracy, and the second change amount calculating means for the measured value observed this time. And a predicted value creating means for creating a predicted value at the time of the next observation. 6. The second unit in the prediction accuracy creating means
The predicted value at the time of measurement is a predicted value created by the predicted value creating means.
The data fluctuation prediction device according to item. 7. The change amount / prediction accuracy detection means, to which the measurement value observed this time belongs, from the measurement value division area change amount and the measurement value division area prediction accuracy stored in the data storage means. After detecting the measured value divided area change amount and measured value divided area prediction accuracy belonging to the measured value divided area, calculate and output each average value of the detected measured value divided area change amount and measured value divided area prediction accuracy, From the number of observations divided region change amount and the number of observations divided region prediction accuracy stored in the data storage unit, the number of observations divided region change amount and the number of observations belonging to the number of observations divided region to which the measurement value observed this time belongs After detecting the divided region prediction accuracy, each average value of the detected number of observations divided region change and the number of observations divided region prediction accuracy is output, and the change rate step change amount stored in the data storage means and From the change rate step prediction accuracy,
After detecting the change rate step change amount and the change rate step prediction accuracy belonging to the change rate step to which the measured value observed this time belongs, output each average value of the detected change rate step change amount and change rate step prediction accuracy. , The change amount step change amount and the change amount step prediction accuracy belonging to the change amount step to which the measured value observed this time belongs from the change amount step change amount and the change amount step prediction accuracy stored in the data storage means. And outputs each average value of the detected change amount step change amount and change amount step prediction accuracy, and the second change amount calculation means divides the measurement value output from the change amount / prediction accuracy detection means. Region change amount, number of observations divided region change amount, change rate step change amount, and change amount step change amount measured value division region prediction accuracy output from the change amount / prediction accuracy detection means, observation Divided area 7. The data variation prediction apparatus according to claim 5, wherein the variation amount is calculated by weighting the average value of the prediction accuracy, the change rate step prediction accuracy, and the change amount step prediction accuracy. 8. The prediction accuracy creating means, for each of the measurement value division area, the observation number division area, the change rate step and the change amount step, prediction accuracy = 1 / | prediction value at the time of the second measurement−the above The measurement value division region prediction accuracy, the observation number division region prediction accuracy, the change rate step prediction accuracy, and the change amount step prediction accuracy are created by the formula of the second measurement value |. The data fluctuation prediction device described in 1. 9. The second change amount calculation means calculates the change amount as follows: change amount = measurement value divided region change amount × measured value divided region prediction accuracy / (measured value divided region prediction accuracy + observation Number division region prediction accuracy + change rate step prediction accuracy + change amount step prediction accuracy) + observation number division region change amount x observation number division region prediction accuracy / (measurement value division region prediction accuracy + observation number division) Region prediction accuracy + change rate step prediction accuracy + change amount step prediction accuracy) + change rate step change amount x change rate step prediction accuracy / (measurement value divided area prediction accuracy + observation number divided area prediction accuracy + Change rate step prediction accuracy + Change amount step prediction accuracy) + Change amount step change amount × Change amount step prediction accuracy / (Measurement value divided area prediction accuracy + Observed number divided area prediction accuracy + Change rate step Prediction accuracy + 9. The data variation prediction device according to claim 5, wherein the variation fluctuation level prediction accuracy) is used for the calculation. 10. The second change amount calculation means calculates the change amount from among the measurement value divided region change amount, the observation number divided region change amount, the change rate step change amount, and the change amount step change amount.
The prediction value corresponding to the largest prediction accuracy among the measurement value divided area prediction accuracy, the observation number divided area prediction accuracy, the change rate step prediction accuracy, and the change amount step prediction accuracy is selected and calculated. The data fluctuation prediction apparatus according to any one of items 1 to 8. 11. A data fluctuation prediction device for predicting future fluctuations of measured values observed in a time series, wherein a range of the measured values is divided at predetermined intervals to create a measured value division area. 1 divided area creating means, a second divided area creating means for creating an observed number divided area by dividing the number of observed objects or the number of observations appearing in time series at predetermined intervals, and observed at the time of the first observation. A change in which the rate of change between the first measured value and the second measured value observed in the second observation after the first observation is divided into predetermined rates to create a change rate step. Rate step creation means and change step creation that creates a change step by dividing the total amount of change from the measurement value observed at the start of observation to each measurement value observed at each observation step by a predetermined ratio Means, the measurement value division area, the observation number division area, the change rate step and The difference between the first measurement value and the second measurement value, the measurement value division area change amount, the number of observations division area change amount, the change rate step change amount, and the change amount step change amount for each change amount step Of the measured value divided area change amount stored in the data storage means and the data storage means for calculating and storing respectively, and belongs to the measured value divided area to which the measured value observed this time belongs and frequency Is detected as the first change amount, and from the observation number division region change amounts stored in the data storage means, the measurement value observed this time belongs to the observation number division region and the frequency is A high change is detected as the second change amount, and from the change rate step change amounts stored in the data storage means, the one that belongs to the change rate step to which the measurement value observed this time belongs and has a high frequency is selected. And the third variation Among the change amount step change amounts stored in the data storage means, one that belongs to the change amount step to which the measurement value observed this time belongs and has a high frequency is detected as the fourth change amount. A change amount detecting means, and an average for calculating an average change amount by averaging the first change amount, the second change amount, the third change amount, and the fourth change amount detected by the change amount detecting means. A change amount calculating means and a predicted value creating means for creating a predicted value for the next observation by adding the average changed amount calculated by the average change amount calculating means to the measured value observed this time Data fluctuation prediction device. 12. A data fluctuation predicting device for predicting future fluctuations of measured values observed in a time series, wherein a range of measured values is divided at predetermined intervals to create a measured value division area. 1 divided area creating means, a second divided area creating means for creating an observed number divided area by dividing the number of observed objects or the number of observations appearing in time series at predetermined intervals, and observed at the time of the first observation. A change in which the rate of change between the first measured value and the second measured value observed in the second observation after the first observation is divided into predetermined rates to create a change rate step. Rate step creation means and change step creation that creates a change step by dividing the total amount of change from the measurement value observed at the start of observation to each measurement value observed at each observation step by a predetermined ratio Means, the measurement value division area, the observation number division area, the change rate step and The difference between the first measurement value and the second measurement value, the measurement value division area change amount, the number of observation times division area change amount, the change rate step change amount, and the change amount step change amount for each change amount step A data storage unit for calculating and storing each of the measurement value division regions, and a measurement value division region belonging to the measurement value division region to which the measurement value observed this time belongs to the measurement value division region change amount stored in the data storage unit. The first change amount is calculated by averaging the change amounts, and among the change numbers of the observation number division regions stored in the data storage means, the observation number division belonging to the observation number division region to which the measurement value observed this time belongs The second change amount is calculated by averaging the region change amounts, and the change rate step change belonging to the change rate step to which the measurement value observed this time belongs among the change rate step change amounts stored in the data storage means. Average the amount of the third A fourth change amount by averaging the change amount step change amounts belonging to the change amount step to which the measurement value observed this time belongs among the change amount step change amounts stored in the data storage means. And a change amount calculating means for calculating, and an average change amount is calculated by averaging the first change amount, the second change amount, the third change amount, and the fourth change amount calculated by the change amount calculating means. Means for calculating the average change amount, and means for creating a predicted value for the next observation by adding the average change amount calculated by the average change amount calculating means to the measured value measured this time A data fluctuation prediction device characterized by: 13. A data fluctuation prediction method for predicting future fluctuations of measured values observed in a time series, wherein a range of possible measured values is divided at predetermined intervals to create a measured value division area, Create the observation number division area by dividing the number of observed objects or the number of observations appearing in time series at predetermined intervals,
The rate of change between the first measurement value observed during the first observation and the second measurement value observed during the second observation after the first observation is divided into steps at predetermined ratios. Areas / stages in which a rate of change step is created and the total amount of change from the measurement value observed at the start of observation to each measurement value observed at each observation time is divided into steps at a predetermined rate Creation step, measurement value division area, observation number division area, change rate step and change amount step, measurement value division area change amount, which is the difference between the first measurement value and the second measurement value A first change amount calculation step for calculating the change amount of the observation number division region, the change amount of the change rate step change amount, and the change amount step change amount; and the measurement value division region, the observation number division region, the change rate step, and the change amount step. For each, the predicted value at the time of the second measurement and the second
Based on the difference from the measured value of
Number of observations divided region prediction accuracy, change rate step prediction accuracy and change amount step prediction accuracy, respectively, and a prediction accuracy creation step, from the measured value divided area change amount, the measurement value to which the measurement value observed this time belongs Those that belong to the divided region and have a high frequency are detected as the first change amount, and from the observed number divided region change amount, those that belong to the observed number divided region to which the measurement value observed this time belongs and have a high frequency Is detected as a second change amount, and from the change rate step change amount,
The third change amount that has a high frequency and belongs to the change rate stage to which the measured value observed this time belongs is detected as a third change amount, and from the change amount stage change amount, the change amount to which the measured value observed this time belongs. A change amount detecting step of detecting a fourth change amount that belongs to a stage and has a high frequency, and the first change amount, the second change amount, the third change amount, and the fourth change amount. A second change amount calculation step of calculating a change amount by weighting the measurement value divided region prediction accuracy, the observation number divided region prediction accuracy, the change rate step prediction accuracy, and the change amount step prediction accuracy; And a predicted value creating step of creating a predicted value for the next observation by adding the amount of change to a measured value. 14. A data fluctuation prediction method for predicting future fluctuations of measured values observed in a time series, wherein a range of the measured values is divided at predetermined intervals to create a measured value division area, Create the observation number division area by dividing the number of observed objects or the number of observations appearing in time series at predetermined intervals,
The rate of change between the first measurement value observed during the first observation and the second measurement value observed during the second observation after the first observation is divided into steps at predetermined ratios. Areas / stages in which a rate of change step is created and the total amount of change from the measurement value observed at the start of observation to each measurement value observed at each observation time is divided into steps at a predetermined rate Creation step, measurement value division area, observation number division area, change rate step and change amount step, measurement value division area change amount, which is the difference between the first measurement value and the second measurement value A first change amount calculation step for calculating the change amount of the observation number division region, the change amount of the change rate step change amount, and the change amount step change amount; and the measurement value division region, the observation number division region, the change rate step, and the change amount step. For each, the predicted value at the time of the second measurement and the second
Based on the difference from the measured value of
Number of observations divided area prediction accuracy, change rate step prediction accuracy and change amount step prediction accuracy, respectively, and a prediction accuracy creating step, and the measurement value division area of the measurement value division area change quantity to which the measurement value observed this time belongs Change amount of the measurement value divided region is calculated by averaging the change amount of the measured value divided region, and the change amount of the observed number divided region which belongs to the observed number divided region to which the measured value observed this time belongs to the change amount of the observed number divided region. The second change amount is calculated by averaging the amounts, and the third change amount is obtained by averaging the change rate step change amounts belonging to the change rate stage to which the measured value observed this time belongs among the change rate step change amounts. An average change amount calculation step of calculating a fourth change amount by averaging change amount step change amounts belonging to the change amount step to which the measurement value observed this time belongs among the change amount step change amounts. The first change amount, the first The change amount, the third change amount, and the fourth change amount are weighted by the measurement value division region prediction accuracy, the observation number division region prediction accuracy, the change rate step prediction accuracy, and the change amount step prediction accuracy. It is characterized by including a second change amount calculation step of calculating a change amount and a predicted value generation step of adding the change amount to the measured value observed this time to create a predicted value for the next observation. Data fluctuation prediction method. 15. The predictive accuracy creating step includes predictive accuracy = 1 / | predicted value at the time of the second measurement−the predictive accuracy for each of the measurement value division area, the observation number division area, the change rate step and the change amount step. 15. The measurement value division region prediction accuracy, the observation number division region prediction accuracy, the change rate step prediction accuracy, and the change amount step prediction accuracy are respectively created by the formula of the second measurement value |. Data fluctuation prediction method described in. 16. The second change amount calculating step calculates the change amount as follows: change amount = first change amount × measured value divided region prediction accuracy / (measured value divided region prediction accuracy + the number of observations) Divided area prediction accuracy + Change rate step prediction accuracy + Change amount step prediction accuracy) + Second change amount x Observation number divided area prediction accuracy / (Measured value divided area prediction accuracy + Observation number divided area prediction) Accuracy + change rate step prediction accuracy + change amount step prediction accuracy) + third change amount x change rate step prediction accuracy / (measurement value divided area prediction accuracy + observation number divided area prediction accuracy + change) Rate step prediction accuracy + Change amount step prediction accuracy) + Fourth change amount x Change amount step prediction accuracy / (Measurement value divided area prediction accuracy + Observation number divided area prediction accuracy + Change rate step prediction accuracy + Precision accuracy of the change level ) The calculation method according to any one of claims 13 to 15, wherein: 17. The second change amount calculation step is performed by changing the change amount to the measured value among the first change amount, the second change amount, the third change amount, and the fourth change amount. 2. The division region prediction accuracy, the number of observations divided region prediction accuracy, the change rate step prediction accuracy, and the change amount step prediction accuracy that correspond to the largest prediction accuracy are selected and calculated.
The data fluctuation amount prediction method according to any one of 3 to 15. 18. A data fluctuation prediction method for predicting future fluctuations of measured values observed in a time series, wherein a range of the measured values is divided at predetermined intervals to create a measured value division area, Create the observation number division area by dividing the number of observed objects or the number of observations appearing in time series at predetermined intervals,
The rate of change between the first measurement value observed during the first observation and the second measurement value observed during the second observation after the first observation is divided into steps at predetermined ratios. Areas / stages in which a rate of change step is created and the total amount of change from the measurement value observed at the start of observation to each measurement value observed at each observation time is divided into steps at a predetermined rate Creation step, measurement value division area, observation number division area, change rate step and change amount step, measurement value division area change amount, which is the difference between the first measurement value and the second measurement value , The number of observations divided region change amount, change rate step change amount and change amount step change amount are respectively calculated, and the measured value divided region change amount, the measurement value to which the measurement value observed this time belongs The one that belongs to the value division area and has a high frequency is detected as the first change amount. Of the changes in the number of observations divided region, the ones that belong to the number of observations divided region to which the measurement value observed this time belongs and have a high frequency are detected as the second changes, and From
The third change amount that has a high frequency and belongs to the change rate stage to which the measured value observed this time belongs is detected as a third change amount, and from the change amount stage change amount, the change amount to which the measured value observed this time belongs. A change amount detecting step of detecting a high frequency item belonging to a stage as a fourth change amount, and averaging the first change amount, the second change amount, the third change amount, and the fourth change amount. And an average change amount calculating means for calculating an average change amount, and a predicted value creating step for adding the average change amount to the measured value observed this time to create a predicted value for the next observation. Data fluctuation prediction method. 19. A data fluctuation prediction method for predicting future fluctuations of measured values observed in a time series, wherein a range of the measured values is divided at predetermined intervals to create a measured value division area, Create the observation number division area by dividing the number of observed objects or the number of observations appearing in time series at predetermined intervals,
The rate of change between the first measurement value observed during the first observation and the second measurement value observed during the second observation after the first observation is divided into steps at predetermined ratios. Areas / stages in which a rate of change step is created and the total amount of change from the measurement value observed at the start of observation to each measurement value observed at each observation time is divided into steps at a predetermined rate Creation step, measurement value division area, observation number division area, change rate step and change amount step, measurement value division area change amount, which is the difference between the first measurement value and the second measurement value A first change amount calculation step of calculating the number of observations divided region change amount, the change rate step change amount, and the change amount step change amount, respectively, and the measurement value observed this time among the measurement value divided region change amount The first change amount obtained by averaging the change amounts of the measurement value division regions belonging to the measurement value division region. A second change amount obtained by averaging the calculated change amounts of the observation number division regions belonging to the observation number division region to which the measurement value observed this time belongs among the observation number division region change amounts stored in the data storage means. Of the change rate step change amounts stored in the data storage means, and the change rate step change amounts belonging to the change rate step to which the measured value observed this time belongs are averaged to calculate a third change amount. Then, among the change amount step change amounts stored in the data storage means, the change amount step change amounts belonging to the change amount step to which the measured value observed this time belongs are averaged to calculate a fourth change amount. A change amount calculation step of 2, and an average change amount calculation step of calculating an average change amount by averaging the first change amount, the second change amount, the third change amount, and the fourth change amount, Add the above average change to the measured value and Data fluctuation prediction method characterized in that and a prediction value generation step of generating a predicted value at the observation.