JP2022034697A - Work support system, and work support method - Google Patents

Abstract

To provide a work support system and a work support method in which a manager who forms a work plan can accurately determine to what degree a predicted value is reliable, and the work plan can be quickly planed, in the work support system and the work support method which support formation of the work plan based on the predicted value of machine learning.SOLUTION: A work support system which supports formation of a work plan comprises: a storage unit which stores a past predicted value and a past actual value, or a past predicted error calculated from the past predicted value and the past actual value; a statistical analysis unit which statistically processes the past predicted error, and finds a predicted error statistical processing result; a predicted error reflection predicted value calculation unit which adds or subtracts the predicted error statistical processing result found by the statistical analysis unit to or from a future predicted value to find a predicted error reflection predicted value; and a display unit which displays the predicted error reflection predicted value found by the predicted error reflection predicted value calculation unit.SELECTED DRAWING: Figure 2

Description

The present invention relates to a business support system and a business support method that support a manager who formulates a business plan based on predicted values of machine learning.

Prediction systems that obtain predicted values using statistical prediction models such as neural networks, deep learning, and multiple regression analysis, which are machine learning, are being put to practical use in various fields. These prediction models give training data in which the values of input items and output items are aligned in advance at the training stage, and adjust the values of the parameters in the prediction model to appropriate values so that the calculated values of the prediction model match as much as possible. .. After that, in the prediction stage, the value of the input item is given to this prediction model, and the prediction value is calculated.

Predictive models are identified to match the training data given to the learning stage as closely as possible, but the predicted values are not always accurate. This becomes remarkable when the quality and quantity of training data are not sufficient. Therefore, it is difficult for the manager who formulates the business plan with reference to the predicted value to judge how much the predicted value can be trusted. In addition, even if it is possible to roughly estimate the prediction error of how much the predicted value deviates, it may not be possible to immediately formulate a business plan such as how to allocate resources accordingly.

As a background technology in this technical field, for example, there is a technology such as Patent Document 1. The sewage inflow amount prediction device of Patent Document 1 includes a storage unit for storing predicted values and prediction errors, and a prediction step width determining unit as components. With this configuration, the time length from the current time to the future time when the sewage inflow is to be predicted can be changed arbitrarily.

Further, in the power prediction method, device, and program of Patent Document 2, a large number of random number sets of power data for each point, power data for each time zone at one point, and a large number for reproducing correlation coefficient values between two time zones are reproduced. It is described that a set of random number data is generated and the predicted values based on them are summarized as a frequency distribution. As a result, the operator of the power system can obtain the predicted range and accuracy of the total power generation output of the region.

Further, Patent Document 3 discloses that in a conventional prediction method using a regression model, a neural network, or the like, an error bar, a standard deviation, or the like is used to evaluate an error in a future predicted value. (Fig. 13 etc.)

Japanese Unexamined Patent Publication No. 2016-21187 Japanese Unexamined Patent Publication No. 2014-54048 Japanese Unexamined Patent Publication No. 2010-20442

However, the above-mentioned Patent Document 1 does not describe the function of showing the user how much the prediction error was. No statistics on the prediction error are given, so the user does not know how likely the prediction is to be off.

Further, since Patent Document 2 is a frequency distribution of predicted values and not an error frequency distribution, the user does not know how likely the predicted values are to deviate.

Further, even when an error bar indicating the maximum / minimum or 3σ is used as in Patent Document 3, the prediction error may be unevenly distributed, and the user must accurately grasp how much the predicted value deviates. I can't.

Therefore, even if the techniques of these patent documents are used, the manager who formulates the business plan cannot judge how much the predicted value can be trusted. Moreover, it is not possible to immediately formulate a business plan based on predicted values, such as how to allocate resources in the future.

Therefore, an object of the present invention is to determine how much the manager who formulates a business plan can trust the predicted value in a business support system and a business support method that support the formulation of a business plan based on the predicted value of machine learning. The purpose is to provide a business support system and a business support method that can make an accurate judgment and can quickly formulate a business plan.

In order to solve the above problems, the present invention is a business support system that supports the formulation of a business plan, in which the past predicted value and the past actual value, or the past predicted value and the past actual value are calculated. A storage unit that stores prediction errors, a statistical analysis unit that statistically processes past prediction errors and obtains prediction error statistical processing results, and a prediction error statistical processing result obtained by the statistical analysis unit are added to or subtracted from future prediction values. It is characterized by including a prediction error reflection prediction value calculation unit for obtaining a prediction error reflection prediction value, and a display unit for displaying the prediction error reflection prediction value obtained by the prediction error reflection prediction value calculation unit.

Further, the present invention is a business support method for supporting the formulation of a business plan, in which (a) a step of calculating a past prediction error from a past predicted value and a past actual value, and (b) the above (a). Prediction error by statistically processing the past prediction error calculated in step and obtaining the prediction error statistical processing result, and (c) adding or subtracting the prediction error statistical processing result obtained in step (b) to the future prediction value. It is characterized by having a step of obtaining a reflected predicted value, and (d) a step of displaying the predicted error reflected predicted value obtained in the step (c) above on the display unit.

According to the present invention, in a business support system and a business support method that support the formulation of a business plan based on the predicted value of machine learning, it is possible to accurately determine how much the manager who formulates the business plan can trust the predicted value. It is possible to realize a business support system and a business support method that can make a judgment and can quickly formulate a business plan.

Issues, configurations and effects other than those described above will be clarified by the following description of the embodiments.

It is a functional block diagram of the business support system which concerns on Example 1 of this invention. This is an example of a display unit that displays a histogram of the prediction error together with the predicted value. This is an example of displaying the frequency (frequency) of the prediction error in one day as a histogram. This is a display example showing the frequency distribution. It is a functional block diagram of the business support system which concerns on Example 2 of this invention. This is an example of a display unit in which a figure (square) that can be freely moved up and down is superimposed and displayed. This is an example of a display unit that displays an automatically calculated future business plan. This is an example of a display unit that displays data that is affected by the automatically calculated future business plan.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. In each drawing, the same components are designated by the same reference numerals, and detailed description of overlapping portions will be omitted.

The business to be supported by the business support system of the present invention specifically includes freight delivery business, plant and equipment operation business, etc., and the future business plan is determined with reference to the predicted values. If it is a business, it is not limited to these.

The business support system and the business support method according to the first embodiment of the present invention will be described with reference to FIGS. 1 to 4.

FIG. 1 is a diagram showing an overall configuration of the business support system of this embodiment. First, the past prediction error 12 stored in the storage unit 10 is given to the statistical analysis unit 14. Further, the learning time storage unit 32, which stores the date and time data when learning is performed by a method such as machine learning in the past, gives the learning time 34 of the prediction model to the specific period setting unit 28.

The learning time 34 of the prediction model is displayed in the specific period setting unit 28, and a screen is displayed on which the administrator can input the specific period setting input value 30 with reference to the learning time 34. The input specific period setting input value 30 is given to the statistical analysis unit 14.

The statistical analysis unit 14 takes out the data of the time corresponding to the specific period setting input value 30 from the past prediction error 12 and statistically analyzes it. The prediction error statistical processing result 16 obtained as a result is given to the prediction error reflection prediction value calculation unit 18. The prediction error reflection prediction value calculation unit 18 is also given a future prediction value 20 predicted by the prediction unit 26. The prediction unit 26 outputs a future prediction value from the reference time.

The prediction error reflection prediction value calculation unit 18 obtains the prediction error reflection prediction value 22 by adding or subtracting the prediction error statistical processing result 16 to the future prediction value 20, and gives it to the display unit 24. The display unit 24 displays the prediction error reflection prediction value 22 on the screen. For example, display a graph in which one axis is time (days) and the other axis is a future prediction error reflection predicted value 22. As a result, the manager can know the predicted value including the statistical value of the past prediction error, and can formulate the business plan more accurately.

Further, the display unit 24 is also given a future predicted value 20, and also displays the future predicted value 20 before addition or subtraction by the prediction error reflection predicted value calculation unit 18. As a result, the manager can know how much the prediction error was, and can accurately judge how reliable the predicted value is.

In the above example, the past prediction error 12 is given to the statistical analysis unit 14 from the storage unit 10 and used as it is, but the storage unit 10 may store the past predicted value and the past actual value, in which case. In the statistical analysis unit 14, the past prediction error 12 may be calculated based on the past predicted value and the past actual value as preprocessing, and the past prediction error 12 may be statistically analyzed.

The actual value 45 stored in the actual value storage unit 43 at a past time point is also given to the display unit 24. Further, the predicted value 36 stored in the predicted value storage unit 42 at the past time point is also given to the display unit 24. The display unit 24 displays the actual value 45 at the past time point and the predicted value 36 at the past time point. This makes it possible to know with what accuracy the future predicted value 20 was correct in the past.

There may be a plurality of predicted values 36 at a past time point depending on how far ahead is predicted even at the same past time point. For example, if there is a forecasting unit 26 that predicts the future forecast value 20 for three periods of 10 minutes, 20 minutes, and 30 minutes as the time length, at the stage of 12:00 yesterday, 12:10, 12:20, Three future forecasts of 20 at 12:30 have been obtained. At 12:10, 10 minutes later, three future forecast values of 20 at 12:20, 12:30, and 12:40 were obtained.

If all of these future predicted values 20 are displayed on the display unit 24, the screen becomes complicated, so it is also good to display only a part of the information. This makes it easier to grasp the information and at the same time reduces the risk of misunderstandings and misunderstandings. The method of fixing the time length is described below as one of the methods of narrowing down the information, but the method of narrowing down is not limited to this.

The target data time length setting unit 48 sets a constant time length 50. The time length 50 set by the target data time length setting unit 48 is given to the display unit 24. For example, if the time length 50 is set to 10 minutes, the display unit 24 displays the future predicted value 20 predicted 10 minutes later at 12:20 at 12:30 yesterday, and 20 at 12:10. Even if there is a future predicted value 20 that predicts minutes later or a future predicted value 20 that predicts 30 minutes later at 12:00, it is not displayed on the screen.

The prediction error statistical processing result 40 at the past time point stored in the prediction error statistical processing result storage unit 38 is given to the past time point prediction error reflection prediction value calculation unit 44. The predicted value calculation unit 44 that reflects the past time point prediction error is also given the predicted value 36 at the past time point from the predicted value storage unit 42. The past time point prediction error reflection prediction value calculation unit 44 obtains the past time point prediction error reflection prediction value 46 by adding or subtracting the past time point prediction error statistical processing result 40 to the past time point prediction value 36, and the display unit 24. Give to. The display unit 24 displays the past time point prediction error reflection prediction value 46 on the screen. This makes it possible to know how the past time point prediction error reflection prediction value 46 at the past time point has changed.

Here, the prediction error statistical processing result 40 at the past time point given to the past time point prediction error reflection prediction value calculation unit 44 is the actual value 45 and the predicted value storage at the past time point stored in the actual value storage unit 43. It may be the result of statistically processing the predicted value 36 stored in the part 42 at the past time point.

FIG. 2 is an example of the display screen of the display unit 24. The downward black triangle in the upper part of the figure indicates the horizontal axis (days) = 0 days, which indicates the current time point. The future predicted value 20 is displayed by a dotted dotted line from the 0th day on the horizontal axis (number of days) to the 4th day. Prediction error reflection The predicted value 22 is displayed as a histogram rotated 90 degrees to the left at the 1st, 2nd, 3rd, and 4th days on the horizontal axis (number of days). On the other hand, the horizontal axis (number of days) is a negative value-4 days to 0 days, the transition from the past to the present is shown as a broken line, the solid line is the actual value 45 at the past time, and the dotted line is. It is a predicted value 36 at a past time point.

In FIG. 2, the histograms displayed at the 1st, 2nd, 3rd, and 4th days on the horizontal axis (days) have many points below the dotted line indicating the future predicted value 20. This means that the actual values obtained in the past so far were lower than the future predicted values 20 obtained by the prediction model (prediction unit 26).

The histogram displayed in FIG. 2 is obtained by the following procedure.

(1) First, the data to be statistically processed to be drawn as a histogram is determined. Therefore, the specific period setting input value 30 is input by the specific period setting unit 28. Here, the period of the data to be analyzed is set. The learning time 34 of the prediction model is displayed in the specific period setting unit 28. For example, the learning time 34 of the previous prediction model is 3 months ago, the learning time 34 of the previous prediction model is 1 year and 3 months ago, and the learning time 34 of the previous prediction model is 2 years and 6 months ago. Is displayed.

The administrator refers to this information and inputs the data to be the target of the histogram in the specific period setting unit 28 from when to when. For example, if the period is set from the past 2 months to 1 month ago, the period of 1 month after the learning time 34 of the previous prediction model can be set as the analysis target.

(2) The statistical analysis unit 14 performs statistical analysis on the past prediction error 12 of the period of the specific period setting input value 30 set in (1). For example, for the histogram displayed at the position where the horizontal axis (number of days) of FIG. 2 is "1 day", the past prediction error 12 that predicts the future by only one day from the present as the time length is used. Table 1 shows an example of frequency distribution data with a past prediction error of 12.

The results of displaying the data in Table 1 as a histogram are shown in FIG. Looking at FIG. 3, it can be seen that the minimum value of the data range of the prediction error of the number of cargoes is -50 [pieces / day] and the maximum value is -10 [pieces / day]. Since the prediction error is about -30 [pieces / day] on average and the frequency around it is high, the future prediction value 20 may deviate from the future prediction value 20 by about -30 [pieces / day] for only one day. Can be inferred to be high. Since the predicted value shown by the dotted line on the horizontal axis (days) in Fig. 2 is 235 [pieces / day], the actual value may be about 235-30 = 205 [pieces / day]. Can be inferred to be high. With the predicted value and FIG. 3 in this way, it is possible to grasp how likely the future predicted value 20 is actually.

In order to grasp this at a glance instead of separate figures, the histogram in FIG. 3 was rotated 90 degrees to the left, and the value of the data section of the prediction error was placed at the position where the future prediction value 20 was added. It is a horizontal histogram inside.

In the histogram of Fig. 3, the part with the highest frequency of -25 [pieces / day] to -35 [pieces / day] has a vertical axis (number of cargoes) of 210 (= 235-25) [pieces / day] in Fig. 2. It is displayed at the position of [day] to 200 (= 235-35) [pieces / day]. Similarly, in the part where the frequency is low -10 [pieces / day] to -15 [pieces / day], the vertical axis (number of cargoes) in Fig. 2 is 225 (= 235-10) [pieces / day] to 220 ( = 235-15) Displayed at the [piece / day] position.

By displaying the histogram rotated 90 degrees to the left in this way, there is a possibility that the number of cargoes will be about 205 (= 235-30) [pieces / day] where the horizontal axis (days) is one day. You can see at a glance what is expensive.

The histogram of the prediction error is not always symmetrical as shown in FIG. 3, and the data section with high frequency is biased to one side instead of the center, for example, as shown at the horizontal axis (days) = 4 days. There may be. If the prediction error is simply displayed by the error bar as in Patent Document 3, such a bias cannot be recognized at all, and there is a possibility that the judgment is erroneous. Therefore, as shown in FIG. 2, by displaying the histogram, the value having a high frequency in the past becomes clear, and a more appropriate judgment can be made.

Since the prediction model trained with training data of insufficient quantity and quality is generally inaccurate, the width of the histogram shown in FIG. 2 in the vertical direction (prediction error) is large, and the deviation from the future prediction value 20 is also large. big. The data gradually increases with the passage of time, and if re-learning is repeated with the training data including it, the accuracy improves, the width of the histogram shown horizontally in FIG. 2 becomes smaller, and the deviation from the future predicted value 20 becomes smaller. Become.

By switching the specific period setting input value 30 after the previous learning, after the previous learning, and after the learning before the previous two times to check the shape and position of the histogram, the administrator can accumulate the learning data. It becomes possible to confirm how much the learning has progressed and the accuracy has improved.

In the period from -4 days to 0 days when the horizontal axis (number of days) in Fig. 2 is a negative value, the actual value 45 at the past time is shown by the solid line, and the dotted line is shown. It is a predicted value 36 at a past time point. The predicted value 36 at the past time point is, for example, a value for which the future is predicted for only one day as the time length at each time point. This figure shows that the actual value 45 at the past time point from -4 days to 0 day was smaller than the predicted value 36 at the past time point.

In addition to these solid and dotted lines, the horizontal axis (days) in Fig. 2 is a negative value-4 days to 0 days, the horizontal axis (days) is the same as 1 to 4 days. The past time prediction error reflection prediction value 46 may be superimposed and displayed on the horizontal histogram. This makes it possible to know how the past time point prediction error reflection prediction value 46 has changed with respect to the past time point prediction value 36.

In the statistical analysis by the statistical analysis unit 14, in addition to the above-mentioned histogram, the maximum value, the minimum value, the average value, the mode value, the median value, and the standard deviation are used as long as the past prediction error 12 is targeted. It may be the method you ask for. All of them are added to or subtracted from the future predicted value 20 and displayed in the figure of FIG. If it is the maximum value and the minimum value, it is displayed as an error bar. If it is the average value, mode value, or median value, it is displayed as one point or one short line. When the standard deviation is used, the value obtained by subtracting the constant multiple of the standard deviation from the mean value and the value obtained by adding the constant multiple of the standard deviation to the mean value are displayed as error bars. The histograms, error bars, single points, or single short lines may not be displayed separately, but may be displayed in duplicate at the same time.

As a diagram showing the frequency distribution displayed on the display unit 24, as shown in FIG. 4, for example, a histogram, a color, a shade of color, a size of a width, a length of a length, a line graph, a curve, an envelope, etc. are displayed. May be.

The prediction unit 26 in FIG. 1 includes at least one of a neural network, deep learning, and multiple regression calculation, but is not limited to these as long as it is a calculation method for machine learning using training data.

As described above, the business support system of this embodiment has a past predicted value (predicted value 36 at a past time point) and a past actual value (actual value 45 at a past time point), or a past predicted value. The storage unit (storage unit 10, predicted value storage unit 42, actual value storage unit 43) that stores the past prediction error 12 calculated from the past actual values and the past prediction error 12 are statistically processed, and the prediction error is predicted. Prediction error reflection predicted value calculation unit that obtains the prediction error reflection prediction value 22 by adding / subtracting the prediction error statistical processing result 16 obtained by the statistical analysis unit 14 and the statistical analysis unit 14 that obtains the statistical processing result 16 to the future prediction value 20. 18 and a display unit 24 for displaying the prediction error reflection prediction value 22 obtained by the prediction error reflection prediction value calculation unit 18.

Further, in the business support method of this embodiment, (a) a past prediction error 12 is obtained from a past predicted value (predicted value 36 at a past time point) and a past actual value (actual value 45 at a past time point). The step to calculate, the step to obtain the prediction error statistical processing result 16 by statistically processing the past prediction error 12 calculated in steps (b) and (a), and the prediction error statistical processing obtained in steps (c) and (b). It has a step of adding and subtracting the result 16 to the future predicted value 20 to obtain a predicted error reflecting predicted value 22, and a step of displaying the predicted error reflected predicted value 22 obtained in steps (d) and (c) on the display unit 24. ing.

As a result, the manager who formulates the business plan can easily guess how much the future forecast value will be after grasping the prediction error of the forecasting department based on the past performance, and a more appropriate business plan. It can be used for the formulation of.

Furthermore, since the manager can interactively and automatically calculate the future business plan by setting the value estimated by himself / herself by moving the figure, the business plan can be formulated in a short time.

The business support system and the business support method according to the second embodiment of the present invention will be described with reference to FIGS. 5 to 8.

FIG. 5 is a diagram showing an overall configuration of the business support system of this embodiment. In the business support system of this embodiment shown in FIG. 5, the graphic position information 52, the business plan automatic calculation unit 54, the business plan 56, and the affected data 58 are added to the components shown in the first embodiment (FIG. 1). ing. Other configurations are the same as those in the first embodiment (FIG. 1). Of these, the figure position information 52 will be described with reference to FIG.

In FIG. 6, squares are displayed on the horizontal axis (number of days) at the 1st, 2nd, 3rd, and 4th days. This figure shows the number of cargoes estimated by the administrator, and can be moved up and down on the display unit 24 on each day by operating the mouse, keyboard, touch panel, and the like. The shape of the figure is not limited to a square, and may be a different shape. The value of the vertical axis (number of cargoes) of the moved figure becomes the figure position information 52.

As shown in FIG. 5, the figure position information 52 is given from the display unit 24 to the business plan automatic calculation unit 54. The business plan automatic calculation unit 54 calculates the business plan 56 based on the graphic position information 52 and gives it to the display unit 24. Further, the data 58 affected by the business plan 56 is also given to the display unit 24 and displayed.

FIG. 7 shows an example of the business plan 56 displayed on the display unit 24. The business plan 56 to be drafted is the number of trucks in operation in this example. FIG. 7 shows how many trucks are required to deliver the cargo to the number of cargoes shown in FIG. The horizontal axis (days) is -4 days to 0 days, and the past actual values including today are displayed, and the horizontal axis (days) is 1 to 4 days in the future business plan 56. Equivalent to.

There are various calculation methods for the business plan 56, but one simple method is to divide the number of cargoes set by the graphic position information 52 by the transport capacity per truck, and round up the value to obtain an integer. To become. It is desirable that FIGS. 6 and 7 are displayed simultaneously on the same screen, and the display of the business plan 56 of FIG. 7 may change in real time according to the operation of moving the figure (square) in FIG. 6 up and down. desirable.

The manager who formulates the business plan estimates the number of cargoes while checking the future predicted value 20 of the number of cargoes and the predicted error reflection predicted value 22 obtained by adding / subtracting the frequency distribution of the past predicted error to the predicted value in FIG. , It is easy to interactively grasp how many trucks will be operated if the estimated number changes.

If it is desirable to have a small number of trucks in operation, and if the number of trucks in operation can be reduced by one by slightly reducing the estimated number, it is easy to formulate a plan under the condition that the estimated number is reduced. On the contrary, if the number of trucks operated increases by one by slightly increasing the estimated number, it is possible to formulate a plan under the condition that the estimated number is increased in consideration of the risk.

FIG. 8 shows a display example of the data 58 affected by the automatically calculated future business plan. As data 58 affected by future business plans, changes in the number of cargoes in stock in the warehouse are displayed in a line graph. Similar to FIG. 7, the horizontal axis (days) is -4 days to 0 days, and the past actual values including today of the inventories in the warehouse are displayed. The data 58 on the horizontal axis (number of days) from 1 to 4 days is affected by the future business plan 56. There are various methods for calculating the affected data 58, but a simple method is the result of subtracting the value obtained by multiplying the number of trucks in FIG. 7 by the transport capacity per truck from the number of cargoes in FIG. It can be calculated as the value obtained by adding the cargo in stock in the warehouse up to the time before that.

It is desirable that FIG. 8 is also displayed at the same time on the same screen as FIG. 6, and the display of the data 58 affected by FIG. 8 changes in real time according to the operation of moving the figure (square) of FIG. 6 up and down. Is desirable.

When the affected data 58 has a plurality of items instead of one item, the trend graph of the affected data 58 for each item may be displayed on the same screen at the same time. In the example of FIG. 8, the warehouse inventory cargo is displayed, but the manager who formulates the business plan can see the change in the number of warehouse inventory cargo while moving the figure (square) in FIG. 6 up and down. With the upper limit of warehouse capacity in mind, it is possible to formulate a plan for the number of trucks in operation so that the number of cargo in stock in the warehouse will be at an appropriate level.

As described above, according to this embodiment, the data affected by the automatically calculated business plan is also displayed. Therefore, in addition to the effect of the first embodiment, the displayed business plan and the affected data are displayed. You can adjust the estimated value while checking.

The present invention is not limited to the above-described embodiment, and includes various modifications. For example, the above embodiments have been described in detail to aid in understanding of the present invention and are not necessarily limited to those comprising all of the described configurations. Further, it is possible to replace a part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. Further, it is possible to add / delete / replace a part of the configuration of each embodiment with another configuration.

10 ... Storage unit 12 ... Past prediction error 14 ... Statistical analysis unit 16 ... Prediction error Statistical processing result 18 ... Prediction error reflection prediction value calculation unit 20 ... Future prediction value 22 ... Prediction error reflection prediction value 24 ... Display unit 26 ... Prediction unit 28 ... Specific period setting unit 30 ... Specific period setting input value 32 ... Learning time storage unit 34 ... Prediction model learning time 36 ... Predicted value at past time 38 ... Prediction error statistical processing result storage unit 40 ... Past Prediction error statistical processing result at time point 42 ... Predicted value storage unit 43 ... Actual value storage unit 44 ... Past time point prediction error reflection predicted value calculation unit 45 ... Actual value at past time point 46 ... Past time point prediction error reflection predicted value 48 … Target data Time length setting unit 50… Time length 52… Graphic position information 54… Business plan automatic calculation unit 56… Business plan 58… Affected data

Claims (15)

In a business support system that supports the formulation of business plans
A storage unit that stores past predicted values and past actual values, or past predicted errors calculated from past predicted values and past actual values.
Statistical analysis unit that statistically processes past prediction errors and obtains prediction error statistical processing results,
Prediction error reflection predicted value calculation unit that obtains the prediction error reflection prediction value by adding / subtracting the prediction error statistical processing result obtained by the statistical analysis unit to the future prediction value.
A display unit that displays the prediction error reflection prediction value obtained by the prediction error reflection prediction value calculation unit, and a display unit.
A business support system characterized by being equipped with. In the business support system according to claim 1,
A business support system characterized by having a prediction unit that outputs future prediction values from the reference time. In the business support system according to claim 2,
The prediction unit is a business support system including at least one of a neural network, deep learning, and multiple regression calculation. In the business support system according to any one of claims 1 to 3,
The display unit is a business support system characterized in that one axis is time and the other axis displays a graph which is a predicted value reflecting a future prediction error. In the business support system according to any one of claims 1 to 4,
A business support system including a specific period setting unit capable of setting a period of past prediction errors to be subject to statistical processing in the statistical analysis unit. In the business support system according to claim 5,
Equipped with a learning time storage unit that stores the prediction model learning time, which is the time when the prediction model learned.
A business support system characterized in that a predictive model learning time output from the learning time storage unit is displayed in the specific period setting unit. In the business support system according to any one of claims 1 to 6,
A business support system characterized in that the prediction error statistical processing result is a diagram showing a frequency distribution. In the business support system according to claim 7,
The diagram showing the frequency distribution is a business support system characterized in that it is shown by at least one of a histogram, a color, a shade of color, a size of a width, a length of a length, a line graph, a curve, and an envelope. In the business support system according to claim 2,
A business support system characterized in that a future predicted value output from the predicted unit is displayed on the display unit. In the business support system according to any one of claims 1 to 9,
The storage unit includes a predicted value storage unit that stores predicted values at a past time point and a predicted value storage unit.
Equipped with an actual value storage unit that stores actual values at the past time,
A business support system characterized in that the predicted value at a past time point output from the predicted value storage unit and the actual value at a past time point output from the actual value storage unit are displayed on the display unit. .. In the business support system according to claim 10,
The storage unit includes a prediction error statistical processing result storage unit that stores prediction error statistical processing results at a past time point.
Prediction error Prediction error reflection prediction by adding or subtracting the prediction error statistical processing result stored in the prediction error statistical processing result storage unit at the past time point and the prediction value at the past time point stored in the prediction value storage unit. It is equipped with a forecast value calculation unit that reflects past time forecast errors to obtain values.
A business support system characterized in that the past time prediction error reflection prediction value obtained by the past time prediction error reflection prediction value calculation unit is displayed on the display unit. In the business support system according to claim 11,
There are multiple forecast values with different time lengths to the future time of the forecast target,
A business support system including a target data time length setting unit for designating a time length of target data used in the past time point prediction error reflection prediction value calculation unit. In the business support system according to any one of claims 1 to 12,
A business plan automatic calculation unit that superimposes a figure that can be moved up and down on the display unit and automatically calculates a future business plan according to the position of the figure.
A screen for displaying a future business plan automatically calculated by the business plan automatic calculation unit, and
A business support system characterized by being equipped with. In the business support system according to claim 13,
A business support system characterized in that data affected by a future business plan automatically calculated by the business plan automatic calculation unit is displayed on the display unit. It is a business support method that supports the formulation of business plans.
(A) Steps to calculate the past prediction error from the past predicted value and the past actual value,
(B) A step of statistically processing the past prediction error calculated in the above (a) step to obtain a prediction error statistical processing result, and a step of obtaining the prediction error statistical processing result.
(C) A step of adding / subtracting the prediction error statistical processing result obtained in the step (b) to the future prediction value to obtain a prediction error reflection predicted value.
(D) A step of displaying the prediction error reflection predicted value obtained in the step (c) above on the display unit, and
A business support method characterized by having.

Images