JPWO2020059099A1 - Label correction device - Google Patents

Abstract

The label correction device includes an input unit for inputting learning data labeled for each block separated by a state change point, which is a point at which the state of time-series data changes, the state change point, and the time. It has a series data, the label, and a correction unit that modifies the label based on the label.

Description

The present invention relates to a label correction device, a label correction method, and a recording medium.

In a plant or factory, time-series data may be acquired from various installed sensors and the system status may be determined from the acquired time-series data. It is known that the state determination of such a system may be performed by using machine learning such as deep learning.

For example, Patent Document 1 is one of the techniques used when determining the state of a system based on time-series data utilizing machine learning. Patent Document 1 describes an information processing device that assigns a label to the measurement data of the sensor. Specifically, the information processing device includes a measurement data acquisition unit, a characteristic data acquisition unit, and a teacher data generation unit. According to Patent Document 1, the characteristic data acquisition unit acquires the characteristic data of the sensor acquired by the maintenance of the sensor. In addition, the teacher data generation unit generates teacher data in which characteristic data is associated with label information with respect to the measurement data of the sensor. According to Patent Document 1, it is possible to reduce the increase in cost associated with maintenance work by the above configuration.

Japanese Unexamined Patent Publication No. 2018-81619

The label given to the measurement data may be incorrect. Therefore, if it is determined that the label is inaccurate, it is desirable to correct the label.

Here, in the case of the time-series data as described above, the system state continues for a certain long period of time, and it is unlikely that the system state changes for each time-series point. Therefore, when modifying the label, it is preferable to consider the characteristics of the time series data as described above. However, there is no known technique for modifying the label in consideration of the characteristics of the time series data. Therefore, there is a problem that the label cannot be corrected while considering the characteristics of the time series data.

Therefore, an object of the present invention is to provide a label correction device, a label correction method, and a recording medium that enable label correction while considering the characteristics of time-series data.

A label correction device according to an embodiment of the present invention for achieving such an object
An input unit for inputting learning data labeled for each block separated by a state change point, which is a point at which the state of time-series data changes.
A correction unit that corrects the label based on the state change point, the time series data, and the label.
It has a structure of having.

In addition, the label correction method, which is another embodiment of the present invention,
The label correction device
Enter the learning data labeled for each block separated by the state change point, which is the point where the state of the time series data changes.
The label is modified based on the state change point, the time series data, and the label.

Moreover, the recording medium which is another form of this invention is
For label correction device,
An input unit for inputting learning data labeled for each block separated by a state change point, which is a point at which the state of time-series data changes.
A correction unit that corrects the label based on the state change point, the time series data, and the label.
It is a computer-readable recording medium on which a program for realizing the above is recorded.

The present invention can provide a label correction device, a label correction method, and a recording medium capable of correcting a label while considering the characteristics of time series data by being configured as described above. It will be possible.

It is a block diagram which shows an example of the structure of the learning apparatus in 1st Embodiment of this invention. It is a figure which shows an example of the time series data acquired by a sensor. It is a figure which shows an example of label information. It is a figure which shows an example of the high reliability period indicated by the high reliability period information. It is a figure which shows an example of the correction difficulty score which the correction difficulty score information shows. It is a figure which shows another example of the correction difficulty score indicated by the correction difficulty score information. It is a figure which shows an example of the correction difficulty score which the correction difficulty score information shows. It is a figure which shows an example of the classification result which the classification result information shows. It is a figure which shows an example of the correction score which the correction score information shows. It is a figure which shows the example of the correction plan at the time of calculating the correction score. It is a figure which shows an example of the display at the time of presenting a correction score. It is a flowchart which shows an example of the operation of a learning device. It is a block diagram which shows an example of the structure of the label correction apparatus.

[First Embodiment]
The first embodiment of the present invention will be described with reference to FIGS. 1 to 12. FIG. 1 is a block diagram showing an example of the configuration of the learning device 1. FIG. 2 is a diagram showing an example of time series data 151 acquired by the sensor 11. FIG. 3 is a diagram showing an example of label information 152. FIG. 4 is a diagram showing an example of the high reliability period indicated by the high reliability period information 153. 5 to 7 are diagrams showing an example of the correction difficulty score indicated by the correction difficulty score information 155. FIG. 8 is a diagram showing an example of the classification result indicated by the classification result information 156. FIG. 9 is a diagram showing an example of the correction score indicated by the correction score information 157. FIG. 10 is a diagram showing an example of the amendment proposal. FIG. 11 is a diagram showing an example of display when presenting a correction score. FIG. 12 is a flowchart showing an example of the operation of the learning device 1.

The learning device 1 (label correction device) monitors various systems such as an industrial plant, a control / control plant, and a factory. For example, the learning device 1 is based on time-series data 151, which is time-series data acquired from one or more sensors 11, and input label information 152, and is used for normal operation, high-load operation, abnormal operation, and stop. Learn the characteristics of data according to the state of the system such as ,. Then, the learning device 1 determines which of the above-mentioned system states the current state corresponds to based on the learning result.

Further, as will be described later, the learning device 1 in the present embodiment has a characteristic of time-series data that the state of the system lasts for a certain long period of time and the state of the system does not change for each time-series point. Based on this, the label information 152 is corrected. Specifically, for example, the learning device 1 changes (for example, moves, adds, deletes, etc.) a system state change point indicating a point (for example, time) at which the system state changes (for example, time) based on a predetermined condition. , The system status label assigned to each block is changed all at once, and the label information 152 is modified. As a result, the learning device 1 corrects the label information 152 while suppressing the possibility of making an unreasonable correction to the time-series data such as the state of the system being changed for each time-series point. In other words, the modification of the learning label information 152 performed by the learning device 1 includes the change of the system state change point and the batch change of the system state label. Further, the change of the system state change point includes the movement of the system state change point, the addition of the system state change point, and the deletion of the system state change point. As will be described later, in the case of the present embodiment, the time series data 151 is divided into a plurality of blocks depending on the system state change point. That is, the block means a part of the time series data 151 separated by the system state change points.

FIG. 1 shows an example of the configuration of the learning device 1. Referring to FIG. 1, for example, the learning device 1 has, as main components, a sensor 11, an operation input unit 12, a screen display unit 13, a communication I / F unit 14, a storage unit 15, and arithmetic processing. It has a part 16.

The sensor 11 acquires time-series data by acquiring data for each time-series point, for example, every minute. The sensor 11 is, for example, a physical sensor such as a pressure gauge, a thermometer, or a flow meter.

In the present embodiment, the number and types of sensors included in the learning device 1 are not particularly limited. The learning device 1 can have, for example, one or more sensors 11 of any number or type.

The operation input unit 12 includes an operation input device such as a keyboard and a mouse. The operation input unit 12 detects the operation of the user who operates the learning device 1 and outputs it to the arithmetic processing unit 16. For example, the user who operates the learning device 1 performs various input operations such as input of label information 152, input of high reliability period information 153, selection of a correction plan to be adopted, and the like, which will be described later, via the operation input unit 12. ..

The screen display unit 13 includes a screen display device such as an LCD (Liquid Crystal Display). The screen display unit 13 displays various information such as time series data 151 on the screen in response to an instruction from the arithmetic processing unit 16.

The communication I / F unit 14 performs data communication with various devices connected via a communication line.

The storage unit 15 is a storage device such as a hard disk or a memory. The storage unit 15 stores processing information and a program 158 necessary for various processes in the arithmetic processing unit 16. The program 158 realizes various processing units by being read and executed by the arithmetic processing unit 16. The program 158 is read in advance from an external device or a storage medium via a data input / output function such as the communication I / F unit 14, and is stored in the storage unit 15. The main information stored in the storage unit 15 is, for example, time series data 151, label information 152, high reliability period information 153, classification model 154, correction difficulty score information 155, and classification result information. There are 156, modified score information 157, program 158, and the like.

The time series data 151 shows the time series data acquired by the sensor 11. In other words, the time series data 151 indicates time series data including a set of data of each time series point acquired by the sensor 11. In other words, the time series data 151 includes data of a plurality of time series points.

In the case of the present embodiment, the learning device 1 has one or more sensors 11. Therefore, for example, when the learning device 1 includes a plurality of sensors 11 of two or more, the time series data 151 includes a plurality of time series data acquired by the plurality of sensors 11.

FIG. 2 shows an example of time series data 151. In the case shown in FIG. 2, the time-series data 151 includes time-series data acquired by the four sensors 11 of the sensor A, the sensor B, the sensor C, and the sensor D. As described above, the time-series data 151 includes time-series data corresponding to the number of sensors 11.

The label information 152 indicates a label given to the time series data 151. In the case of the present embodiment, the label information 152 is composed of a system state change point (state change point) and a system state label (label) given to each block separated by the system state change point. .. As will be described later, the label information 152 is input by, for example, an operation by a user who operates the learning device 1. The label information 152 may be configured so that the label information input unit 162 of the learning device 1 automatically inputs the label information 152 based on the time series data 151 or the like.

Here, the system state change point means a point where the state of time-series data changes. The system state conversion point indicates a point (for example, time) at which the state of the system monitored by the sensor 11 changes. The time series data 151 is divided into a plurality of blocks according to the system state change point. Further, the system status label is a label given to each block separated by the system status change point, and indicates the system status. That is, in the case of the present embodiment, the same system state label is given to the data of the time series points belonging to the same block. The system status label includes, for example, system status A (normal operation), system status B (high load operation), system status C (abnormal operation), system status D (stop), and the like. The system state label is used when learning is performed by the classification model learning unit 164, which will be described later, or when the label information 152 is corrected by the correction unit 167. The system status label may be other than the one illustrated above, such as determining abnormal operation in more detail.

FIG. 3 shows an example of label information 152 given to the time series data 151 shown in FIG. For example, in the case of FIG. 3, the time series data 151 is separated by four system state change points. Further, in the case shown in FIG. 3, the system state A (normal state) and the system state B (high) are sequentially arranged from the left to the right in FIG. 3 for the five blocks separated by the four system state change points. Four types of system status labels are assigned: load operation), system status A (normal operation), system status C (abnormal operation), and system status D (stopped). In other words, for example, the time series points belonging to the leftmost block shown in FIG. 3 are given a system state label of system state A (normal operation).

As described above, the label information 152 is composed of a system state change point for dividing the time series data 151 and a system state label given to each block divided by the system state change point. The number of system state change points and the number and types of system state labels included in the label information 152 may be other than those illustrated above.

The high-reliability period information 153 indicates a high-reliability period, which is a period of time-series data 151 that is judged to be highly reliable (for example, a period that is considered to have a low possibility of incorrect labeling). There is. As will be described later, the first learning is performed based on the system state label of the time series points within the high reliability period indicated by the high reliability period information 153. Therefore, it is desirable that the high reliability period be set for all types of system state labels included in the label information 152 described above. That is, in the case of the above-mentioned example, the reliability is high for all types of system state A (normal state), system state B (high load operation), system state C (abnormal operation), and system state D (stop). It is desirable that a period is set.

As will be described later, the high reliability period is input by, for example, an operation by a user who operates the learning device 1. The high reliability period may be input by the high reliability period information input unit 163 of the learning device 1 based on the distance from the system state change point or the like.

FIG. 4 shows an example of the high reliability period input for the time series data 151 shown in FIG. In FIG. 4, a high reliability period is input for a predetermined period during the system state A (normal operation) and the system state D (stop). In the case shown in FIG. 4, the high reliability period during the system state A (normal operation) and the system state D (stop) is separated from the system state change point by, for example, a predetermined value (any value may be used) or more. ing. In FIG. 4, the high reliability period is not input for the system state B and the system state C. Therefore, for the system state B and the system state C, it is desirable that the high reliability period is input for some period in the time series data (not shown in FIG. 4).

The classification model 154 is a model generated as a result of learning by the classification model learning unit 164. As will be described later, the classification model learning unit 164 performs learning for the high reliability period indicated by the high reliability period information 153. Therefore, it can be said that the classification model 154 is a model generated as a result of learning for a high reliability period. In other words, the classification model 154 is a model generated as a result of learning on a part of the time series data 151.

The correction difficulty score 155 indicates a correction difficulty score which is a value indicating the difficulty when correcting the label. The correction difficulty score is calculated by the correction difficulty score calculation unit 165, for example, for time-series points that have not been learned by the classification model learning unit 164.

The correction difficulty score indicates, for example, that the larger the value, the more difficult it is to correct the system state change point or the system state label. As will be described later, for example, the correction difficulty score can be calculated based on the time series data and the system state change point. Specifically, for example, the correction difficulty score is classified into the same block as the time-series points and the degree of distance from the system state change point of each time-series point constituting the time-series data. It can be calculated based on whether or not the period during which it was performed is included. As will be described later, the correction difficulty score is also used when calculating the correction score as a guide when adopting the correction plan. Therefore, it can be said that the correction difficulty score is a value for correcting the result of classification by the classification unit 166, which will be described later, when the label information 152 is corrected.

FIG. 5 shows a system state label included in the label information 152, a learned flag indicating whether or not learning has been performed by the classification model learning unit 164 (that is, whether or not it is within the high reliability period), and a correction. It is a figure which shows an example of correspondence with a difficulty score. FIG. 5 illustrates a case where there is a system state change point between the time series points “12:04” and “12:05”.

In the case shown in FIG. 5, the classification model learning unit 164 has already learned the data of the time series points at the times of "12:00" and "12:01". Therefore, it is impossible to correct the system state label, and the correction difficulty score calculation unit 165 does not calculate the correction difficulty score. As a result, the correction difficulty score is "(cannot be corrected)". As described above, the correction difficulty score is not calculated for the time series points learned by the classification model learning unit 164 (that is, the time series points within the high reliability period).

Further, in the case of FIG. 5, the correction difficulty score corresponding to "12:02" is "0.8", the correction difficulty score corresponding to "12:03" is "0.7", and ". The correction difficulty score corresponding to "12:04" is "0.6". Similarly, in the case of FIG. 5, the correction difficulty score corresponding to “12:05” is “0.2”, and the correction difficulty score corresponding to “12:06” is “0.3”. The correction difficulty score corresponding to "12:07" is "0.4", and the correction difficulty score corresponding to "12:08" is "0.5".

As described above, in the case of FIG. 5, the correction difficulty score becomes a larger value as the distance from the system state change point increases. Further, the correction difficulty score is a large value when the learned flag belongs to a block including a period in which the learned flag is round (that is, learning by the classification model learning unit 164 has already been performed). The detailed calculation method of the correction difficulty score will be described later.

The time-series data is divided into blocks according to the system state change point. Therefore, as shown in FIG. 6, each time-series point in the block may have system state change points at both ends. In such a case, the correction difficulty score can be calculated based on, for example, the degree of distance from the closer system state change point. For example, in the case shown in FIG. 6, the correction difficulty score corresponding to “12:07” is “0.4”, and the correction difficulty score corresponding to “12:08” is “0.3”. The correction difficulty score corresponding to "12:09" is "0.2".

FIG. 7 shows an example of the tendency of the correction difficulty score when the correction difficulty score is calculated for the time series data 151 shown in FIG. With reference to FIG. 7, it can be seen that the correction difficulty score becomes a larger value as the distance from the system state change point increases. Further, the correction difficulty score becomes a large value when it belongs to the block including the period in which the high reliability period information 153 is input (that is, the learning by the classification model learning unit 164 has already been performed). You can see that there is.

The classification result information 156 shows the result of classification by the classification unit 166. As will be described later, the classification unit 166 describes the time series points that were not used for learning by the classification model learning unit 164, the characteristics of the time series point data, and the classification model 154 generated by the classification model learning unit 164. Based on this, the time series points are classified. Therefore, the classification result information 156 shows the result of classification for the time series points that were not used for learning by the classification model learning unit 164.

FIG. 8 is a diagram showing an example when the classification result information 156 is added to FIG. Referring to FIG. 8, “12:00” and “12:01” are time-series points used for learning by the classification model learning unit 164, and therefore do not show the result of classification. Further, referring to FIG. 8, for example, the classification result corresponding to "12:02" is "system state A", the classification result for "12:03" is "system state A", and "12:04". The classification result for "" is "system state B". Further, the classification result for "12:05" is "system state C", the classification result for "12:06" is "system state C", and the classification result for "12:07" is "system state B". The classification result for "12:08" is "system state C".

In the case shown in FIG. 8, for the time series points of "12:02", "12:03", and "12:07", the system status label first assigned by the label information input unit 162 and the classification unit 166 are used. There is no difference from the classification result. On the other hand, time-series points other than the above, such as "12:04", "12:05", "12:06", and "12:08", are first assigned by the label information input unit 162. There is a difference between the system status label given and the result of classification by the classification unit 166. As described above, there may be a discrepancy between the system state label first input by the user or the like and the result of classification by the classification unit 166.

The modified score information 157 indicates the modified score. The correction score is a score indicating the degree to which the classification is correct after correction and the correction is easy. The amendment score can be used as a guide for determining whether or not to adopt the amendment proposal. The correction score is calculated by the correction unit 167 using the result of classification by the classification unit 166 and the correction difficulty score.

The correction score indicates, for example, that the larger the value, the less the discrepancy between the classification result by the classification unit 166 and the system state label actually given by the correction. In addition, the correction score indicates that the larger the value, the easier and more rational the correction is. For example, the correction score is based on the total number of time-series points that are correctly classified by the correction, the total number of time-series points that are incorrectly classified by the correction, and the total number of target correction difficulty scores. Can be calculated.

FIG. 9 shows an example of the value of the correction score for each correction plan. Further, FIG. 10 shows a specific example of the modification assumed by each modification plan. For example, referring to FIG. 10, in "Modification 1", the system state change point is moved from "12:04" and "12:05" to "12:03" and "12:04". Let me. With reference to FIG. 9, the correction score when the above-mentioned revision plan 1 is performed is, for example, "0.4". Further, referring to FIG. 10, in "Modification 2", the system state change point is moved from "12:04" and "12:05" to "12:02" and "12:03". Let me. With reference to FIG. 9, the correction score when the above-mentioned correction plan 1 is performed is, for example, "-1.3". As described above, the correction score information 157 includes the correction score when each correction proposal is made.

The arithmetic processing unit 16 has a microprocessor such as an MPU and its peripheral circuits, and by reading and executing the program 158 from the storage unit 15, the hardware and the program 158 cooperate to realize various processing units. do. As the main processing units realized by the arithmetic processing unit 16, for example, a data acquisition unit 161, a label information input unit 162, a high reliability information period input unit 163, a classification model learning unit 164, and a correction difficulty score. There are a calculation unit 165, a classification unit 166, and a correction unit 167.

The data acquisition unit 161 acquires the data of each time series point acquired by the sensor 11. Then, the data acquisition unit 161 stores the data acquired from the sensor 11 as time-series data 151 in the storage unit 15.

The label information input unit 162 is an input unit that accepts the input of the label information 152. Then, the label information input unit 162 stores the received label information 152 in the storage unit 15. As described above, the label information 152 that receives the input by the label information input unit 162 includes the system state change point and the system state label.

For example, it is assumed that the time series data 151 as shown in FIG. 2 is displayed on the screen display unit 13. In such a case, the label information input unit 162 receives an input by the user via the operation input unit 12 for the time series data 151 displayed on the screen display unit 13, for example, to set the system state change point and the system. It accepts the input of the status label and the label information 152 including. In this way, the label information input unit 162 accepts the input of the label information 152 by, for example, accepting the user's input for the time series data 151 displayed on the screen display unit 13. As a result, the label information 152 is displayed on the screen display unit 13 together with the time series data 151, for example, as shown in FIG.

Alternatively, for example, the label information input unit 162 automatically accepts the input of the label information 152 based on the time series data 151 or the like. In the present embodiment, the process for automatically accepting the input of the label information 152 is not particularly limited. The label information input unit 162 can accept the input of the label information 152 by using a known method.

It is desirable that the label information 152 input by the label information input unit 162 is based on the characteristics of the time series data. Therefore, for example, a predetermined period is provided between the system state change points. In other words, for example, the label information input unit 162 includes a label between the system state change point and the system state change point so as to include a time series point equal to or higher than a predetermined reference (any value may be used). Accepts the input of information 152.

The high reliability period information input unit 163 accepts the input of the high reliability period. Then, the high reliability period information input unit 163 stores the high reliability period information 153 indicating the received high reliability period in the storage unit 15.

For example, it is assumed that the time series data 151 as shown in FIG. 2 is displayed on the screen display unit 13. Alternatively, it is assumed that the time series data 151 and the label information 152 are displayed on the screen display unit 13 as shown in FIG. In such a case, the high reliability period information input unit 163 receives an input by the user via the operation input unit 12 for the time series data 151 displayed on the screen display unit 13, for example, so that the high reliability period information input unit 163 has a high reliability period. Accepts input. At this time, the high reliability period may straddle the system state change point. In this way, the high-reliability period information input unit 163 accepts the input of the high-reliability period by, for example, accepting the user's input for the time-series data 151 displayed on the screen display unit 13.

Alternatively, for example, the high-reliability period information input unit 163 automatically accepts the input of the high-reliability period based on the degree of distance from the system state change point. For example, the high-reliability period information input unit 163 defines a high-reliability period as a period that is separated from the system state change point by an arbitrary value or more. The high-reliability period information input unit 163 includes the number of time-series points within the high-reliability period for each system status label, the ratio of the high-reliability period to the high-reliability period for each system status label, and the period outside the high-reliability period. , The high reliability period may be determined by referring to information other than the degree of distance from the system state change point.

The classification model learning unit 164 is a learning unit that performs learning for the high reliability period indicated by the high reliability period information 153. For example, the classification model learning unit 164 generates a classification model 154 by deep learning based on the characteristics of the data of the time series points belonging to the high reliability period and the system state label given to the time series points. do. After that, the classification model learning unit 164 stores the generated classification model 154 in the storage unit 15. In this way, the classification model learning unit 164 generates the classification model 154 by performing learning for the high reliability period, which is a part of the time series data indicated by the time series data 151.

In this embodiment, the processing when the classification model learning unit 164 performs learning is not particularly limited. For example, the classification model learning unit 164 can perform learning using known means such as learning using a neural network.

The correction difficulty score calculation unit 165 calculates the correction difficulty score for the time-series points that have not been learned by the classification model learning unit 164. Then, the correction difficulty score calculation unit 165 stores the calculated correction difficulty score as the correction difficulty score information 155 in the storage unit 15.

For example, the correction difficulty score calculation unit 165 learns the degree of distance from the system state change point of each time series point constituting the time series data 151 and the classification model learning unit 164 into the same block as the time series point. The correction difficulty score is calculated based on whether or not the period is included. For example, the correction difficulty score calculation unit 165 is +0.4 when there is a learned time-series point in the same block, +0 when there is no learned time-series point, and +0. By performing the calculation of 1, the correction difficulty score of each time series point to be calculated is calculated. As described above, each time-series point in the block may have system state change points at both ends. In such a case, the correction difficulty score calculation unit 165 can calculate the correction difficulty score based on the system state change point closer to the time series point.

Specifically, for example, in the case of FIG. 5, the correction difficulty score calculation unit 165 has learned time series points in the same block with respect to the time series points of “12:02”, so that the time series points are +0. 4. In addition, the total of +0.4 "0.8" is calculated as the correction difficulty score according to the degree of distance from the system state change point. Similarly, in the case shown in FIG. 5, for example, the correction difficulty score calculation unit 165 is +0 with respect to the time series point of "12:05" because there is no learned time series point in the same block. The total of +0.2 "0.2" is calculated as the correction difficulty score according to the degree of distance from the system state change point. Further, when the system state change points are provided at both ends as shown in FIG. 6, the correction difficulty score calculation unit 165 has already learned the time series points of, for example, "12:08" in the same block. +0 because there is no time-series point of, and +0.3 in total "0.3" according to the degree of separation based on the system state change point on the right side of FIG. 6, which is closer to the time-series point of "12:08". Is calculated as the correction difficulty score.

In this way, the correction difficulty score calculation unit 165 is divided into the same blocks as the time-series points and the degree of separation of the data of each time-series point constituting the time-series data from the system state change point by the model learning unit 164. The correction difficulty score is calculated based on whether or not the learning period is included. In other words, the correction difficulty score calculation unit 165 calculates the correction difficulty score from the viewpoint of whether or not the system state label can be easily changed and whether or not the system state change point can be easily changed. .. That is, if there are trained time series points in the same block, it will be difficult to modify the system state label. Therefore, the correction difficulty score calculation unit 165 raises the correction difficulty score when there are learned time-series points in the same block. Further, as the time series point is farther from the system state change point, it is necessary to move the system state change point to a large extent, and it becomes difficult to move the system state change point. Therefore, the correction difficulty score calculation unit 165 adds the correction difficulty score according to the degree of distance from the system state change point.

The calculation method by the correction difficulty score calculation unit 165 is just an example. The correction difficulty score calculation unit 165 may calculate the correction difficulty score by using a method other than the above-exemplified method.

The classification unit 166 describes the time series points that were not used for learning by the classification model learning unit 164 based on the characteristics of the data of the time series points and the classification model 154 generated by the classification model learning unit 164. Classify the series points. That is, the classification unit 166 classifies the time-series points in the period other than the predetermined period based on the classification model 154 generated as a result of learning for the predetermined period in the time-series data 151. For example, in the case of the present embodiment, the classification unit 166 sets the states of the time series points to system state A (normal operation), system state B (high load operation), system state C (abnormal operation), and system state D (stop). ). The classification unit 166 may perform classification other than those illustrated above.

The classification unit 166 classifies based on the classification model 154 without considering the characteristics of the time series data. Therefore, the result of classification by the classification unit 166 may be unreasonable as a classification for time-series data, for example, the system state may be changed for each time-series point (see FIG. 8). .. Therefore, in the present embodiment, the result of classification by the classification unit 166 is not used as it is for modifying the label information 152. By going through the processing by the correction unit 167, which will be described later, it is possible to make corrections based on the characteristics of the time series data.

The correction unit 167 calculates a correction score for each correction proposal. Then, the correction unit 167 stores the calculated correction score as the correction score information 157 in the storage unit 15. Further, the correction unit 167 displays the calculated correction score on the screen display unit 13. Then, the correction unit 167 determines the correction plan to be adopted according to the input by the user. After that, the correction unit 167 reflects the revised proposal adopted. That is, the correction unit 167 changes the label information 152 so as to include the adopted correction proposal.

For example, the correction unit 167 moves the system state change point, adds or deletes the system state change point, changes the system state label given to the block at once, the above combination, and the like. Is calculated.

Specifically, in the case shown in FIG. 10, in the amendment plan 1 for which the amendment score is calculated, the system state change point between "12:04" and "12:05" is set to "12:03". It is a plan to move it between "12:04". Further, the amendment 5 is a plan to collectively change the system state label given to the block on the right side of the system state change point from the system state B (high load operation) to the system state C (abnormal operation). In addition, the amendment 6 adds a system state change point between "12:07" and "12:08", and adds a system state change point between "12:04" and "12:05". It is a plan to change the system state label of the block between the system state change point and the system state change point to system state C (abnormal operation). As described above, the correction proposals for which the correction score is calculated include changes in the system state change points, batch changes in the system state labels, and combinations thereof.

Further, the correction score is calculated by the correction unit 167 based on, for example, the following formula.
Total number of time series points for correct classification-Total time series points for incorrect classification-Total difficulty score for correction

For example, referring to FIGS. 9 and 10, in the case of the above-mentioned amendment 1, the block to which "12:04" belongs by moving the system state change point between "12:03" and "12:04". Fluctuates. As a result, the system status label given to "12:04" changes from system status A (normal operation) to system status B (high load operation). As a result, the system state label input by the label information input unit 162 and the classification result by the classification unit 166 are correct. Therefore, the correction unit 167 calculates 1-0-0.6 = 0.4 as the correction score. Further, in the case of the amendment 5, the system status labels given to "12:05", "12:06", "12:07", and "12:08" are changed from the system status B (high load operation) to the system status. It changes to C (abnormal operation). As a result, the system status labels of "12:05", "12:06", and "12:08" are correct, but the system status labels of "12:07" are incorrect. Therefore, the correction unit 167 calculates 3-1 (0.2 + 0.3 + 0.4 + 0.5) = 0.6 as the correction score. Similarly, in the case of the amendment 6, the system status labels given to "12:05", "12:06", and "12:07" are changed from system status B (high load operation) to system status C (abnormal operation). Changes to. As a result, the system status labels of "12:05" and "12:06" are correct, but the system status labels of "12:07" are incorrect. Therefore, the correction unit 167 calculates 2-1 (0.2 + 0.3 + 0.4) = 0.1 as the correction score.

For example, as described above, the correction unit 167 calculates correction scores for a plurality of correction proposals. Then, the correction unit 167 causes the screen display unit 13 to display the calculated correction score. At this time, as shown in FIG. 11, for example, the correction unit 167 rearranges the calculated correction score in the ranking format and displays it on the screen display unit 13. For example, in FIG. 11, the one with the higher correction score is rearranged so as to be located higher. After that, the correction unit 167 receives input from the user via the operation input unit 12 or the like, and determines the correction plan to be adopted. Then, the correction unit 167 reflects the revised proposal adopted. That is, the correction unit 167 updates the label information 152. Note that the correction unit 167 may perform a display other than the above example, for example, displaying a correction score in a ranking format together with a visualization of the correction plan as shown in FIG. Further, the correction unit 167 may be configured to determine the correction score to be adopted based on a predetermined standard. For example, the correction unit 167 has a predetermined threshold value (any value may be used). Then, the correction unit 167 determines as a correction plan to adopt the correction plan having a correction score equal to or higher than a predetermined threshold value.

Note that the correction unit 167 sets the correction score calculation target only for correction proposals in which the size of the blocks separated by the system state change points (for example, the number of time-series points included in the blocks) does not fall below a predetermined threshold value. It is desirable to do. For example, when the system state change point is moved or newly added, the size of the block becomes excessively small, and there is a possibility that the correction plan ignores the characteristics of the time series data. By selecting the amendment plan based on the size of the block as described above, the amendment unit 167 can calculate the amendment score based on the amendment plan with the above-mentioned risk reduced.

In addition, the period of the amendment proposed by the user is highly reliable. Therefore, the high reliability period information input unit 163 can add the period of the amendment proposed by the user to the high reliability period. That is, the high-reliability period information input unit 163 can update the high-reliability period information 153 in response to the result of the correction by the correction unit 167. In addition, the classification model learning unit 164 can perform learning for a high reliability period including a newly added period. Further, based on the result, the classification unit 166 performs a new classification by the newly learned classification model 154. As described above, the learning device 1 can be configured to perform the process reflecting the result of the correction by the correction unit 167 again.

The above is the description of the configuration of the learning device 1.

Next, the operation of the learning device 1 will be described. FIG. 12 shows an example of the operation of the learning device 1.

Referring to FIG. 12, the data acquisition unit 161 acquires the data of each time series point acquired by the sensor 11. Then, the data acquisition unit 161 stores the data acquired from the sensor 11 as time-series data 151 in the storage unit 15. Further, the time series data 151 acquired by the data acquisition unit 161 is displayed on the screen display unit 13 (step S101).

The label information input unit 162, for example, receives an input by the user via the operation input unit 12 for the time series data 151 displayed on the screen display unit 13, so that the system state change point and the system state label can be set. The input of the label information 152 including the label information 152 is accepted (step S102). Alternatively, the label information input unit 162 automatically accepts the input of the label information 152 based on the time series data 151 or the like.

The high reliability period information input unit 163 accepts the input of the high reliability period information 153 (step S103). The high reliability period information 153 is also input, for example, by accepting an input by the user via the operation input unit 12 for the time series data 151 displayed on the screen display unit 13. Alternatively, the high-reliability period information input unit 163 automatically accepts the input of the high-reliability period information 153 based on, for example, the degree of distance from the system state change point.

The classification model learning unit 164 performs learning for the period in which the high reliability period information 153 is input (step S104). For example, the classification model learning unit 164 generates the classification model 154 by deep learning based on the characteristics of the data of the time series points belonging to the above period and the system state label given to the time series points.

The correction difficulty score calculation unit 165 calculates the correction difficulty score for the time-series points that were not learned by the classification model learning unit 164 (step S105). Further, the classification unit 166 describes the time-series points that were not used for learning by the classification model learning unit 164 based on the characteristics of the data of the time-series points and the classification model 154 generated by the classification model learning unit 164. , The time series points are classified (step S106).

Either the process of step S105 or the process of step S106 may be performed first. Further, the process of step S105 and the process of step S106 may be performed in parallel.

The correction unit 167 calculates a correction score for each correction plan (step S107). Then, the correction unit 167 displays the calculated correction score on the screen display unit 13. At this time, the correction unit 167 displays the calculated correction score on the screen display unit 13 in a ranking format, for example (step S108). Further, the correction unit 167 accepts the input of the correction proposal to be adopted, which is the selection result from the user input via the operation input unit 12 or the like (step S109). In response to this, the correction unit 167 reflects the accepted adoption proposal.

When the input of the amendment proposal adopted by the amendment unit 167 is received (step S110, Yes), the high reliability period information input unit 163 adds the period of the amendment proposal adopted by the user to the high reliability period information 153. .. Further, the classification model learning unit 164 performs learning for the period in which the high reliability period information 153 including the newly added period is input (step S111). After that, the process returns to step S105. After that, the same processing as the above-mentioned processing is performed. On the other hand, when the input of the correction plan adopted by the correction unit 167 is not accepted (step S110, No), the process ends.

The above is an example of the operation of the learning device 1.

As described above, the learning device 1 in the present embodiment includes the label information input unit 162, the high reliability period information input unit 163, the classification model learning unit 164, the correction difficulty score calculation unit 165, and the classification unit 166. , And a correction unit 167. With such a configuration, the classification model learning unit 164 can learn only during the period in which the system state label input by the high reliability period information input unit 163 is determined to have high reliability. In addition, the classification unit 166 can classify the time series points that were not used for learning by the classification model learning unit 164. As a result, the correction unit 167 corrects the system state label input by the label information input unit 162 based on the classification result by the classification unit 166 and the calculation result by the correction difficulty score calculation unit 165. Can be done. As a result, it is possible to correct the system state label given in the period with low reliability based on the learning result in the period with high reliability in the time series data 151.

Further, in the case of the present embodiment, the correction unit 167 moves the system state change point, adds or deletes the system state change point, and collectively changes the system state label given to the block. It is configured to calculate the correction score for each correction plan such as. Then, the correction unit 167 is configured to reflect the correction plan selected by the user among the above-mentioned correction plans. With such a configuration, it is possible to reduce the risk of making corrections that do not take into account the characteristics of the time-series data, such as changing the system state for each time-series point. As a result, it is possible to correct the system state label while considering the characteristics of the time series data.

Further, in the case of the present embodiment, the correction unit 167 is configured to calculate the correction score in consideration of the correction difficulty score. With such a configuration, the user can determine the correction plan to be adopted based on the correction score in consideration of the correction difficulty score. Here, the correction difficulty score is the difficulty level when correcting the system state label, and is a standard for judging whether or not the change of the system state label is rational. Therefore, with the above configuration, it is possible to more accurately present the correction score, which is a reference when the user makes a rational judgment.

The learning device 1 does not necessarily have to be composed of one information processing device. The learning device 1 may be realized by a plurality of information processing devices having the above-mentioned functions.

Further, in the present embodiment, the learning device 1 monitors various systems such as an industrial plant, a control / control plant, and a factory. However, the application destination of the learning device 1 is not limited to the case of handling physical data such as the system illustrated above. The learning device 1 can be applied to a whole system having characteristics of time-series data, such as less possibility that the state of the system is changed for each time-series point. The sensor 11 does not necessarily have to be a physical sensor.

Further, in the present embodiment, it is assumed that the learning device 1 has the sensor 11. However, the learning device 1 does not necessarily have to have the sensor 11. For example, the data acquisition unit 161 may be configured to acquire the time series data 151 from the sensor 11 outside the learning device 1 via a network or the like.

Further, in the present embodiment, the learning device 1 uses the correction difficulty score. However, the learning device 1 does not necessarily have to use the correction difficulty score. That is, the learning device 1 does not have to have the correction difficulty score calculation unit 165. In this case, the correction unit 167 can be configured to calculate the correction score based on, for example, the total number of time-series points for which the classification is correct and the total number of time-series points for which the classification is incorrect.

[Second Embodiment]
Next, a second embodiment of the present invention will be described with reference to FIG. FIG. 13 is a block diagram showing an example of the configuration of the label correction device 2.

Referring to FIG. 13, the label correction device 2 has an input unit 21 and a correction unit 22. For example, the label correction device 2 has a storage device for storing a program and an arithmetic unit such as a CPU (Central Processing Unit). For example, the label correction device 2 realizes the processing unit by executing the program stored in the storage device by the arithmetic unit.

The input unit 21 inputs learning data labeled for each block separated by a state change point, which is a point at which the state of time-series data changes.

The correction unit 22 corrects the label based on the state change point, the time series data, and the label.

As described above, the label correction device 2 has an input unit 21 and a correction unit 22. With such a configuration, the label correction device 2 can change the label by the correction unit 22. According to such a configuration, since the label assigned to each block is corrected, the label can be corrected in consideration of the characteristics of the time series data.

Further, the label correction device 2 described above can be realized by incorporating a predetermined program into the label correction device 2 which is a computer. Specifically, in a program according to another embodiment of the present invention, a label is assigned to the label correction device 2 for each block separated by a state change point, which is a point at which the state of time-series data changes. This is a program for realizing an input unit 21 for inputting data, a state change point, time-series data, and a correction unit 22 for correcting a label based on a label.

Further, in the label correction method executed by the label correction device 2 described above, the label correction device 2 assigns a label to each block separated by a state change point, which is a point at which the state of time-series data changes. It is a method of inputting training data and correcting the label based on the state change point, the time series data, and the label.

Even in the invention of the program, the label correction method, or the computer-readable recording medium having the above-described configuration, in order to have the same operation and effect as the label correction device 2. , The above-mentioned object of the present invention can be achieved.

<Additional notes>
Part or all of the above embodiments may also be described as in the appendix below. Hereinafter, the outline of the label correction device and the like in the present invention will be described. However, the present invention is not limited to the following configurations.

(Appendix 1)
An input unit for inputting learning data labeled for each block separated by a state change point, which is a point at which the state of time-series data changes.
A correction unit that corrects the label based on the state change point, the time series data, and the label.
Label correction device with.
(Appendix 2)
The label correction device according to Appendix 1.
A learning unit that generates a classification model by learning for a part of the time series data,
Using the classification model, a classification unit that classifies the period that is not the target of learning by the learning unit, and a classification unit.
Have,
The correction unit is a label correction device that corrects the label based on the state change point and the result of classification by the classification unit.
(Appendix 3)
The label correction device described in Appendix 2,
It has a calculation unit that calculates a correction difficulty score indicating the difficulty level when correcting the label.
The correction unit corrects the label by using the result of classification by the classification unit and the correction difficulty score.
Label correction device.
(Appendix 4)
The label correction device described in Appendix 3,
The calculation unit is a label correction device that calculates the correction difficulty score based on the degree of distance from the state change point and the range in which learning by the learning unit is performed.
(Appendix 5)
The label correction device according to Appendix 3 or Appendix 4.
From the number that the result of classification by the classification unit and the label are correct by changing the state change point, the correction unit is based on the result of classification by the classification unit and the input unit by modifying the label. By subtracting the number of incorrect labels and the correction difficulty score, a correction score indicating the degree to which the classification is correct after correction and the degree of ease of correction is calculated, and the correction score is calculated. Modify the label based on
Label correction device.
(Appendix 6)
The label correction device according to any one of Supplementary note 1 to Supplementary note 5.
The correction unit is a label correction device that corrects the label by changing the state change point.
(Appendix 7)
The label correction device according to Appendix 6.
The correction unit is a label correction device that changes the state change point by moving the state change point.
(Appendix 8)
The label correction device according to Appendix 6 or Appendix 7.
The correction unit is a label correction device that changes the state change point by adding or deleting the state change point.
(Appendix 9)
The label correction device according to any one of Supplementary note 1 to Supplementary note 8.
The correction unit is a label correction device that corrects the label by collectively changing the label given to the block.
(Appendix 10)
The label correction device
Enter the learning data labeled for each block separated by the state change point, which is the point where the state of the time series data changes.
A label correction method for modifying the label based on the state change point, the time series data, and the label.
(Appendix 11)
For label correction device,
An input unit for inputting learning data labeled for each block separated by a state change point, which is a point at which the state of time-series data changes.
A correction unit that corrects the label based on the state change point, the time series data, and the label.
A computer-readable recording medium that records a program to achieve this.

The programs described in each of the above embodiments and appendices may be stored in a storage device or recorded in a computer-readable recording medium. For example, the recording medium is a portable medium such as a flexible disk, an optical disk, a magneto-optical disk, and a semiconductor memory.

Although the present invention has been described above with reference to each of the above embodiments, the present invention is not limited to the above-described embodiments. Various changes that can be understood by those skilled in the art can be made to the structure and details of the present invention within the scope of the present invention.

1 Learning device 11 Sensor 12 Operation input unit 13 Screen display unit 14 Communication I / F unit 15 Storage unit 151 Time series data 152 Label information 153 High reliability period information 154 Classification model 155 Correction difficulty score information 156 Classification result information 157 Correction Score information 158 Program 16 Arithmetic processing unit 161 Data acquisition unit 162 Label information input unit 163 High reliability period information input unit 164 Classification model learning unit 165 Correction difficulty score calculation unit 166 Classification unit 167 Correction unit 2 Label correction device 21 Input unit 22 Correction part

Claims (11)

An input unit for inputting learning data labeled for each block separated by a state change point, which is a point at which the state of time-series data changes.
A correction unit that corrects the label based on the state change point, the time series data, and the label.
Label correction device with. The label correction device according to claim 1.
A learning unit that generates a classification model by learning for a part of the time series data,
Using the classification model, a classification unit that classifies the period that is not the target of learning by the learning unit, and a classification unit.
Have,
The correction unit is a label correction device that corrects the label based on the state change point and the result of classification by the classification unit. The label correction device according to claim 2.
It has a calculation unit that calculates a correction difficulty score indicating the difficulty level when correcting the label.
The correction unit corrects the label by using the result of classification by the classification unit and the correction difficulty score.
Label correction device. The label correction device according to claim 3.
The calculation unit is a label correction device that calculates the correction difficulty score based on the degree of distance from the state change point and the range in which learning by the learning unit is performed. The label correction device according to claim 3 or 4.
From the number that the result of classification by the classification unit and the label are correct by changing the state change point, the correction unit is based on the result of classification by the classification unit and the input unit by modifying the label. By subtracting the number of incorrect labels and the correction difficulty score, a correction score indicating the degree to which the classification is correct after correction and the degree of ease of correction is calculated, and the correction score is calculated. Modify the label based on
Label correction device. The label correcting device according to any one of claims 1 to 5.
The correction unit is a label correction device that corrects the label by changing the state change point. The label correction device according to claim 6.
The correction unit is a label correction device that changes the state change point by moving the state change point. The label correction device according to claim 6 or 7.
The correction unit is a label correction device that changes the state change point by adding or deleting the state change point. The label correcting device according to any one of claims 1 to 8.
The correction unit is a label correction device that corrects the label by collectively changing the label given to the block. The label correction device
Enter the learning data labeled for each block separated by the state change point, which is the point where the state of the time series data changes.
A label correction method for modifying the label based on the state change point, the time series data, and the label. For label correction device,
An input unit for inputting learning data labeled for each block separated by a state change point, which is a point at which the state of time-series data changes.
A correction unit that corrects the label based on the state change point, the time series data, and the label.
A computer-readable recording medium that records a program to achieve this.

Images