JP7406447B2 - Computer system and item demand forecasting method - Google Patents

Description

The present invention relates to systems and methods for predicting demand for items.

A method of predicting the demand for objects to be traded is to input input data such as transaction history into a mathematical model (prediction model) such as linear regression or neural network to predict future demand for the object. is known (for example, see Patent Document 1).

Patent Document 1 states, ``A demand forecasting method for predicting future demand is based on a prediction model that is constructed based on actual measured values of demand and explanatory variables that are external factors that influence increases or decreases in demand. A step of calculating the error between the predicted value of the past demand amount calculated by inputting the actual measured value of the explanatory variable and the actual measured value of the past demand amount, and whether the calculated error corresponds to an abnormal value. a step of determining whether the error corresponds to an abnormal value, a step of obtaining a new explanatory variable, a step of updating the prediction model based on the obtained new explanatory variable, and a step of updating the prediction model based on the obtained new explanatory variable. and a step of newly predicting the future demand amount using the predicted model.

Generally, predictive models are generated by machine learning using learning data. When the amount of training data showing a certain feature is small, the feature is ignored in the prediction model. That is, the feature is not reflected in the prediction model. Therefore, predictive models have difficulty accurately predicting demand for objects during busy periods. This is because sudden changes in object demand do not occur often, and the amount of learning data regarding object demand during busy periods is sufficiently smaller than the amount of learning data regarding object demand during normal periods. On the other hand, a technique described in Patent Document 2 is known as a method for predicting demand during busy periods.

Patent Document 2 states, ``The power demand peak prediction device 1 includes a past data storage unit 12 that stores past data in which the amount of electricity used at each time in the past day is associated with characteristic information of the past day, and a prediction target. a feature information acquisition unit 11 that obtains feature information of the day; a past data extraction unit 13 that extracts one or more past data containing feature information that is the same as or similar to the feature information of the prediction target day from the past data storage unit 12; For each past data extracted by the past data extraction unit 13, a peak identifying unit 14 identifies a power demand peak value or a power demand peak time period, and based on the power demand peak value or power demand peak time period of each past data, , a prediction unit 15 that calculates a predicted value of a power demand peak value or a power demand peak time period on a prediction target day, and an output unit 16 that outputs the predicted value calculated by the prediction unit 15.” has been done.

JP 2017-16632 Publication Japanese Patent Application Publication No. 2015-139283

As described in Patent Document 2, a predictive model generated by machine learning using past data as learning data strongly depends on the learning data. On the other hand, in many fields such as logistics, demand peaks occur periodically, but often at different times.

Therefore, with conventional prediction models, it is difficult to predict demand for items that takes into account deviations in busy seasons that do not appear in training data. This is because conventional prediction models treat the timing of the peak season as fixed.

An object of the present invention is to realize a system and method that can specify a busy season from input data and predict demand for items during the busy season and other periods with high accuracy.

A typical example of the invention disclosed in this application is as follows. That is, the computer system includes at least one computer and predicts the demand for an item in a prediction period, the computer system includes history information that stores history regarding shipments of each of a plurality of items per first unit period, and demand for the item. definition information of a prediction model for predicting the prediction period, a busy season calculation unit that calculates a busy season that is a period during which the shipment amount of the first item is greater than a threshold; definition information of the prediction model, the history of the first item, and the busy season. a forecasting unit that generates , as demand forecast information, time-series data indicating the predicted shipment amount of the first item per second unit period of the forecast period , based on the forecast period; and an item per third unit period of the forecast period. a peak distribution calculation unit that calculates a peak distribution that is a distribution of the number of times a peak in which a shipment amount of A peak distribution is calculated, and a second peak distribution is calculated using the history of the second item, and the busy season calculation unit calculates a mixed peak distribution based on the first peak distribution and the second peak distribution. and calculate the busy season based on the mixed peak distribution, and the prediction unit generates the demand forecast information using the definition information of the prediction model and the history of the first item, The demand forecast information is updated by correcting the time series data included in the busy season using the second peak distribution, and the updated demand forecast information is output .

According to the present invention, it is possible to identify the busy season for an item and predict the demand for the item in a prediction period including the busy season with high accuracy. Problems, configurations, and effects other than those described above will be made clear by the description of the following examples.

1 is a diagram illustrating an example of a configuration of a demand forecasting system according to a first embodiment; FIG. FIG. 3 is a diagram showing an example of a data structure of shipping history management information according to the first embodiment. FIG. 3 is a diagram illustrating details of a functional unit of a computer according to the first embodiment. 3 is a flowchart illustrating an overview of processing executed by the prediction unit of Example 1. FIG. 5 is a flowchart illustrating an example of a peak distribution calculation process executed by the prediction unit of the first embodiment. 3 is a diagram showing an example of a data structure of peak distribution information in Example 1. FIG. 3 is a diagram showing an example of a data structure of peak distribution information in Example 1. FIG. 3 is a diagram showing an example of a graph visualizing the peak distribution of Example 1. FIG. 3 is a diagram showing an example of a graph visualizing the peak distribution of Example 1. FIG. 7 is a flowchart illustrating an example of a busy season calculation process executed by the prediction unit of the first embodiment. FIG. 3 is a diagram showing an example of a data structure of busy season information according to the first embodiment. 3 is a flowchart illustrating an example of a demand forecasting process executed by the forecasting unit of the first embodiment. FIG. 3 is a diagram showing an example of the data structure of demand forecast information according to the first embodiment. 7 is a flowchart illustrating an example of a prediction correction process executed by the prediction unit according to the first embodiment. FIG. 3 is a diagram showing an example of a demand forecast screen displayed by the display unit of the first embodiment. 7 is a diagram illustrating an example of a correction basis screen displayed by the display unit of Example 1. FIG. 2 is a diagram illustrating an example of the configuration of a demand forecasting system according to a second embodiment. FIG. FIG. 3 is a diagram illustrating details of a functional unit of a computer according to a second embodiment. 7 is a flowchart illustrating an example of a demand forecasting process executed by the forecasting unit according to the second embodiment.

Embodiments of the present invention will be described below with reference to the drawings. However, the present invention should not be construed as being limited to the contents described in the Examples shown below. Those skilled in the art will readily understand that the specific configuration can be changed without departing from the spirit or spirit of the present invention. In the configuration of the invention described below, the same or similar configurations or functions are denoted by the same reference numerals, and redundant explanations will be omitted. In this specification, etc., expressions such as "first," "second," and "third" are used to identify constituent elements, and do not necessarily limit the number or order. The position, size, shape, range, etc. of each component shown in the drawings etc. may not represent the actual position, size, shape, range, etc. in order to facilitate understanding of the invention. Therefore, the present invention is not limited to the position, size, shape, range, etc. disclosed in the drawings and the like.

FIG. 1 is a diagram illustrating an example of the configuration of a demand forecasting system according to a first embodiment.

The demand forecasting system is a computer system composed of at least one computer 100. The computer 100 includes a CPU 110, an IO interface 111, a network interface 112, a main storage device 113, and a secondary storage device 114. Each piece of hardware is connected to each other via a bus.

CPU 110 executes programs stored in main storage device 113. By executing processing according to a program, the CPU operates as a functional unit (module) that implements a specific function. In the following description, when a process is explained using a functional unit as a subject, it is meant that the CPU executes a program that implements the functional unit.

The IO interface 111 is an interface for connecting to an external device. The computer 100 is connected to an input device 101 and an output device 102 via an IO interface 111. The input device 101 is a keyboard, a mouse, a touch panel, or the like. The output device 102 is a display, a printer, or the like.

Network interface 112 is an interface for connecting to an external device via network 103. The network 103 is a WAN (Wide Area Network), a LAN (Local Area Network), or the like. The connection format of the network 103 may be either wired or wireless.

The main storage device 113 stores programs executed by the CPU 110 and data used by the programs. The main storage device 113 also includes a work area that is temporarily used by the program. The main storage device 113 is, for example, a DRAM (Dynamic Random Access Memory).

The main storage device 113 stores programs that implement the input section 120, the prediction section 121, and the display section 122. Note that the main storage device 113 stores programs (not shown) such as an OS (Operating System), but these are omitted in this embodiment.

The input unit 120 receives various inputs from the user via the input device 101 or the network 103. The prediction unit 121 predicts the demand for the item in the prediction period and outputs the prediction result. A forecast period is a time range over which demand for an item is predicted, such as years, months, and weeks. Note that the forecast period is not limited to these. It is assumed that the prediction period is set in advance. The display unit 122 presents the prediction results and the like to the user via the output device 102. Note that the computer 100 accepts input from the user by receiving data from a user terminal communicably connected via the network 103, and transmits output to be output by the output device 102 to the user terminal. By doing so, data may be output to the user terminal.

The secondary storage device 114 permanently stores data. The secondary storage device 114 is, for example, an HDD (Hard Disk Drive) or an SSD (Solid State Drive).

The secondary storage device 114 stores shipping history management information 130 and predictive model management information 131. Note that the secondary storage device 114 stores information (not shown), such as information for managing the characteristics of items, but is omitted in this embodiment.

The shipping history management information 130 stores data regarding past item shipments (shipping history). The shipping history includes the amount of items shipped during the measurement period. The measurement period is, for example, a day, a week, a month, or the like. Note that the measurement period is not limited to these. The data structure of the shipping history management information 130 will be explained using FIG. 2. The predictive model management information 131 stores definition information of a mathematical model (predictive model) that predicts demand for items from shipping history. The prediction model is, for example, a decision tree, a neural network, or the like. Note that the predictive model management information 131 may store different model definition information for each item characteristic, such as item type and item category. Note that the present invention is not limited to the type and structure of the prediction model.

Note that the demand forecasting system may be composed of a plurality of computers 100. In this case, functional units are distributed and arranged in a plurality of computers 100. For example, the input section 120 is arranged in one computer 100, the prediction section 121 is arranged in one computer 100, and the display section 122 is arranged in one computer 100.

Note that the programs and data stored in the main storage device 113 may be stored in the secondary storage device 114. In this case, the CPU 110 reads the program and data from the secondary storage device 114 and stores them in the main storage device 113.

FIG. 2 is a diagram showing an example of the data structure of the shipping history management information 130 according to the first embodiment.

The shipping history management information 130 stores the shipping history of multiple items. Specifically, the shipping history management information 130 stores entries including an item ID 201, a shipping date 202, and a shipping amount 203. One entry corresponds to one shipping history. The shipping history management information 130 shown in FIG. 2 stores shipping history whose measurement period is one day (first unit period).

Item ID 201 is a field that stores item identification information. The shipping date 202 is a field that stores the shipping date of the item. The shipping amount 203 is a field that stores the shipping amount of the item.

FIG. 3 is a diagram illustrating details of the functional units of the computer 100 according to the first embodiment.

The input unit 120 receives a request to output a demand forecast for the target item. The output request includes identification information of the target item.

The input unit 120 identifies related items that have a relationship with the target item based on the characteristics of the target item. For example, an item that is similar or identical to the target item in at least one of the category, storage location (warehouse), manufacturer, and shipping destination is specified as a related item. Note that the viewpoint of item relevance is not limited to these. Note that the number of related items to be specified may be one, or two or more.

The input unit 120 inputs, from the shipping history management information 130, a dataset consisting of a shipping history group of the target item in an arbitrary processing period, and a dataset consisting of a shipping history group of related items in an arbitrary processing period. read out. The input unit 120 inputs the shipping history information 310 and 311 storing the read data set to the prediction unit 121. The shipping history information 310 stores a plurality of data sets (for example, a group of shipping histories of target items for three years), and the shipping history information 311 stores a plurality of data sets (for example, a group of shipping history of related items for three years). shipping history group) are stored. The processing period is a time range of the shipping history used by the prediction unit 121, and is, for example, a year, a month, and a week. Note that the present invention is not limited to the processing period. It is assumed that the processing period is set in advance. In this example, it is assumed that the processing period and the prediction period are the same.

The prediction unit 121 uses the shipping history information 310 of the target item and the shipping history information 311 of related items to predict the demand (shipment amount) of the target item in the prediction period, and outputs the prediction result to the display unit 122.

Here, details of the prediction unit 121 will be explained. The prediction unit 121 includes a peak distribution calculation unit 300, a busy season calculation unit 301, a demand prediction unit 302, and a correction unit 303.

The peak distribution calculation unit 300 calculates the peak distribution of the target item using the shipping history information 310 of the target item, and outputs the calculation result as peak distribution information 320. The peak distribution calculation unit 300 also calculates the peak distribution of the related item using the shipping history information 311 of the related item, and outputs the calculation result as peak distribution information 321.

Here, the peak distribution is a distribution of the number of times a peak occurs during the processing period. The peak distribution can be used as information indicating the trend of demand for an item. A peak represents a state in which the amount of items shipped per unit period is greater than the demand threshold. The unit period is a period that is smaller than the processing period and larger than the measurement period, and is, for example, a month or a week. Note that the unit period is not limited to these. It is assumed that the unit period is set in advance.

The busy season calculation unit 301 calculates the busy season using the peak distribution information 320 and 321, and outputs the calculation result as the busy season information 330.

The demand prediction unit 302 predicts the demand (predicted shipment amount) of the target item in the prediction period based on the prediction model management information 131 and the shipment history information 310 of the target item, and outputs the prediction result as demand prediction information 340.

The correction unit 303 uses the peak distribution information 321 and the busy season information 330 to correct the predicted shipment amount of the target item stored in the demand forecast information 340, and outputs the correction result as the demand forecast information 350.

The display unit 122 presents information regarding the demand forecast for the target item based on the demand forecast information 350 output from the forecast unit 121.

Regarding each functional unit that the computer 100 has, a plurality of functional units may be combined into one functional unit, or one functional unit may be divided into a plurality of functional units for each function. For example, the demand forecasting unit 302 may have the function of the correction unit 303.

First, the processing executed by the prediction unit 121 will be described using FIGS. 4 to 12.

FIG. 4 is a flowchart outlining the processing executed by the prediction unit 121 of the first embodiment.

The prediction unit 121 starts processing when the shipping history information 310, 311 is input from the input unit 120. Note that the prediction unit 121 stores the input shipping history information 310 and 311 in the work area. First, the prediction unit 121 executes a peak distribution calculation process (step S101). Through this processing, peak distribution information 320, 321 is generated. Details of the peak distribution calculation process will be explained using FIG. 5.

Next, the prediction unit 121 executes a busy season calculation process (step S102). Through this process, busy season information 330 is generated. Details of the busy season calculation process will be explained using FIG. 8.

Next, the prediction unit 121 executes a demand prediction process (step S103). Through this process, demand forecast information 340 is generated. Details of the demand prediction process will be explained using FIG. 10.

Next, the prediction unit 121 executes prediction correction processing (step S104). Through this process, demand forecast information 350 is generated. Details of the prediction correction process will be explained using FIG. 12.

Next, the prediction unit 121 outputs the demand prediction information 350 to the display unit 122 (step S105), and ends the process.

FIG. 5 is a flowchart illustrating an example of the peak distribution calculation process executed by the prediction unit 121 of the first embodiment. 6A and 6B are diagrams illustrating an example of a data structure of peak distribution information according to the first embodiment. 7A and 7B are diagrams showing examples of graphs visualizing the peak distribution of Example 1.

The peak distribution calculation unit 300 acquires the shipping history information 310 of the target item from the work area (step S201).

Next, the peak distribution calculation unit 300 generates demand time series data by totaling the shipping history group of each data set included in the shipping history information 310 for each unit period (step S202).

For example, the peak distribution calculation unit 300 generates demand time-series data by aggregating the shipping history that stores the daily shipping amount included in the shipping history information 310 on a weekly basis according to the shipping date chronologically. . Identification information indicating the order of the time series in the processing period is given to the demand time series data. When the processing period is one year, numbers from 1 to 52 are assigned to the demand time series data as identification information. Demand time series data is generated for each dataset.

Next, the peak distribution calculation unit 300 selects one data set from the plurality of data sets (step S203). Here, it is assumed that data sets are selected in chronological order.

Next, the peak distribution calculation unit 300 identifies a peak in the processing period based on the demand time series data of the selected data set (step S204).

Specifically, the peak distribution calculation unit 300 compares the total value of the shipping amount of the shipping history included in the demand time series data (shipping amount of items per week) with the demand threshold value, and calculates the total value of the shipping amount. If it is larger than the demand threshold, it is determined that a peak has occurred. The peak distribution calculation unit 300 stores aggregated data made up of identification information of demand time series data identified as a peak in a work area.

The demand threshold is, for example, the sum of the average shipping amount of the item and the standard deviation of the shipping amount of the item. However, the demand threshold is not limited to the above.

Next, the peak distribution calculation unit 300 determines whether processing of all data sets has been completed (step S205).

If the processing of all data sets has not been completed, the peak distribution calculation unit 300 returns to step S203 and executes the same processing.

When the processing of all datasets is completed, the peak distribution calculation unit 300 calculates the distribution of the number of peak occurrences in the processing period (peak distribution) based on the aggregated data stored in the work area (step S206). , and output as peak distribution information 320. Peak distribution information 320 is stored in the work area.

For example, peak distribution information 320 as shown in FIG. 6A is output. The peak distribution information 320 stores entries consisting of a time series number 601 and a number of peak occurrences 602. One entry corresponds to the calculation result of the number of peak occurrences in one unit period. The time series number 601 is a field that stores identification information of demand time series data. The number of peak occurrences 602 is a field that stores the number of peak occurrences.

The peak distribution information 320 can be expressed as a graph as shown in FIG. 7A. The horizontal axis in FIG. 7A represents the time series of the processing period, and the vertical axis represents the number of peak occurrences.

Next, the peak distribution calculation unit 300 acquires the shipping history information 311 of the related item from the work area (step S207).

Next, the peak distribution calculation unit 300 generates demand time series data by totaling the shipping history group of each data set included in the shipping history information 311 for each unit period (step S208). The process in step S208 is the same as the process in step S202.

Next, the peak distribution calculation unit 300 selects one data set from the plurality of data sets (step S209). Here, it is assumed that data sets are selected in chronological order.

Next, the peak distribution calculation unit 300 identifies a peak in the processing period based on the demand time series data of the selected data set (step S210). The process in step S210 is the same as the process in step S204.

Next, the peak distribution calculation unit 300 determines whether processing of all data sets has been completed (step S211).

If the processing of all data sets has not been completed, the peak distribution calculation unit 300 returns to step S209 and executes the same processing.

When the processing of all data sets is completed, the peak distribution calculation unit 300 calculates the distribution of the number of peak occurrences in the processing period (peak distribution) based on the aggregated data stored in the work area (step S212). , and output as peak distribution information 321. Peak distribution information 321 is stored in the work area.

For example, peak distribution information 321 as shown in FIG. 6B is output. The peak distribution information 321 has the same data structure as the peak distribution information 320.

The peak distribution information 321 can be expressed as a graph as shown in FIG. 7B. The horizontal axis in FIG. 7B represents the time series of the processing period, and the vertical axis represents the number of peak occurrences.

Note that when there are multiple related items, the peak distribution calculation unit 300 executes the processes from step S207 to step S211 for each related item, and calculates the peak distribution of each related item. Further, the peak distribution calculation unit 300 outputs a distribution obtained by superimposing the peak distributions of each related item as peak distribution information 321.

FIG. 8 is a flowchart illustrating an example of the busy season calculation process executed by the prediction unit 121 of the first embodiment. FIG. 9 is a diagram showing an example of the data structure of the busy season information 330 according to the first embodiment.

The busy season calculation unit 301 acquires the peak distribution information 320 of the target item and the peak distribution information 321 of related items from the work area (step S301).

Next, the busy season calculation unit 301 calculates a mixed peak distribution using the peak distribution information 320 and 321 (step S302).

For example, the busy season calculation unit 301 calculates a mixed peak distribution by calculating the average value of the number of times a peak occurs in each unit period using each peak distribution. Note that the method for calculating the mixed peak distribution is not limited to the above-mentioned method.

Next, the busy season calculation unit 301 calculates a busy season threshold based on the mixed peak distribution (step S303).

For example, the busy season calculation unit 301 calculates the average value of the number of peak occurrences in the mixed peak distribution as the busy season threshold. Note that the method of calculating the busy season threshold is not limited to the above-mentioned method.

Next, the busy season calculation unit 301 searches for a busy season in the processing period based on the mixed peak distribution and the busy season threshold (step S304).

Specifically, the busy season calculating unit 301 searches for an interval (a set of demand time series data) in which the number of peak occurrences is greater than the busy season threshold and the time series is continuous in the mixed peak distribution. This section is during the busy season. The busy season calculation unit 301 stores search data including the searched section and identification information in the work area. When multiple sections are searched, multiple pieces of search data are stored in the work area.

Next, the busy season calculation unit 301 determines whether there is at least one busy season (step S305).

If there is no busy season, the busy season calculation unit 301 ends the busy season calculation process.

If at least one busy season exists, the busy season calculation unit 301 generates busy season information 330 (step S306). Thereafter, the busy season calculation unit 301 ends the busy season calculation process.

For example, busy season information 330 as shown in FIG. 9 is output. The busy season information 330 stores entries including an item ID 901, a busy season ID 902, and a section 903. One entry corresponds to one busy season. Item ID 901 is a field that stores identification information of a target item. The busy season ID 902 is a field that stores identification information of the busy season. A section 903 is a field that stores information about a section corresponding to a busy season. Section 903 stores information on a section specified using a time-series identification number.

By using the peak distribution of the target item and related items, it is possible to determine the busy season of the target item, taking into account the demand for the related item. This allows for liquidity during busy periods.

FIG. 10 is a flowchart illustrating an example of the demand prediction process executed by the prediction unit 121 of the first embodiment. FIG. 11 is a diagram showing an example of the data structure of the demand forecast information 340 according to the first embodiment.

The demand forecasting unit 302 acquires the shipping history information 310 of the target item from the work area (step S401).

Next, the demand forecasting unit 302 generates demand time-series data by aggregating the shipping history included in the shipping history information 310 for each unit period (step S402). The processing in step S402 is the same as the processing in step S202.

Next, the demand forecasting unit 302 uses the demand time series data to calculate feature amounts to be input into the forecasting model (step S403).

For example, the demand forecasting unit 302 calculates the year, season, week, etc. of shipment as feature quantities. Note that the type of feature amount to be calculated and the method for calculating the feature amount are not limited.

Next, the demand prediction unit 302 acquires the prediction model management information 131 (step S404), and predicts the demand for the target item in the prediction period using the prediction model management information 131 and the feature amount (step S405).

Specifically, the demand forecasting unit 302 constructs a forecasting model defined in the forecasting model management information 131, inputs feature amounts to the forecasting model, and outputs a forecast result of the demand for the target item. For example, only 52 pieces of prediction data indicating the predicted shipment amount of the target item per week (second unit period) are output at weekly intervals. The plurality of prediction data are output as time series data.

Next, the demand prediction unit 302 generates demand prediction information 340 based on the prediction result (step S406), and stores it in the work area.

For example, demand forecast information 340 as shown in FIG. 11 is generated. The demand forecast information 340 stores entries including an item ID 1101, a date 1102, and a shipping amount 1103. One entry corresponds to forecast data indicating the predicted shipment amount of the target item for a certain week of the processing period. Item ID 1101 is a field that stores identification information of the target item. Date 1102 is a field that stores the date and time that specifies the predicted week. In this embodiment, the predicted weekly shipment amount is calculated. Therefore, the date of a specific day of the week is set to date 1102. The shipping amount 1103 is a field that stores the predicted shipping amount of the target item.

FIG. 12 is a flowchart illustrating an example of the prediction correction process executed by the prediction unit 121 of the first embodiment.

The correction unit 303 determines whether a busy season exists (step S501).

Specifically, the correction unit 303 determines whether busy season information 330 is stored in the work area. If the busy season information 330 is stored in the work area, the correction unit 303 determines that a busy season exists.

If there is no busy season, the correction unit 303 ends the prediction correction process. In this case, uncorrected demand forecast information 340 is output to display section 122.

If a busy season exists, the correction unit 303 acquires demand forecast information 340 from the work area (step S502).

Next, the correction unit 303 acquires peak distribution information 321 of related items from the work area (step S503).

Next, the correction unit 303 updates the demand forecast information 340 based on the busy season and peak distribution information 321 (step S504). After that, the correction unit 303 ends the prediction correction process. Specifically, the following processing is executed.

(Process A1) The correction unit 303 selects a target busy season from among the busy seasons managed by the busy season information 330.

(Process A2) The correction unit 303 identifies, from the demand forecast information 350, forecast data in which the date 1102 is included in the target busy season.

(Process A3) The correction unit 303 corrects the shipment amount 1103 of the specified predicted data based on the number of times a peak occurs during the busy season in the peak distribution of related items managed in the peak distribution information 321. For example, the correction unit 303 increases the value of the shipping amount 1103 according to the number of times the peak occurs during the busy season. The amount of increase can be set arbitrarily. Note that the correction method is not limited to this.

(Processing A4) The correction unit 303 determines whether processing has been completed for all busy seasons. If the processing has not been completed for all busy periods, the correction unit 303 returns to processing A1 and executes the same processing. When the processing has been completed for all busy seasons, the correction unit 303 stores the demand forecast information 350 in which the forecast data for the busy seasons has been corrected in the work area.

Next, the process executed by the display unit 122 will be described using FIGS. 13 and 14.

FIG. 13 is a diagram illustrating an example of a demand prediction screen 1300 displayed by the display unit 122 of the first embodiment. FIG. 14 is a diagram illustrating an example of a correction basis screen 1400 displayed by the display unit 122 of the first embodiment.

The display unit 122 acquires demand forecast information 340 , busy season information 330 , peak distribution information 320 , and shipping history information 310 along with demand forecast information 350 from the forecast unit 121 .

Display unit 122 generates display information for displaying demand forecast screen 1300 based on demand forecast information 350, demand forecast information 340, busy season information 330, peak distribution information 321, and shipping history information 310.

The demand forecast screen 1300 is a screen for presenting demand forecasts, and includes a demand forecast display field 1310 and a forecast basis display button 1320.

The demand forecast display column 1310 is a column that displays a graph showing the demand forecast for the target item. The horizontal axis of the graph represents time-series identification information of prediction data during the processing period, and the vertical axis represents shipping amount.

The demand forecast display column 1310 in FIG. 13 displays a graph that connects demand forecasts for a plurality of processing periods. The processing period is one year, and the unit period is one week. In this case, the processing period is divided into 52 unit periods, and each unit period is given identification information from 1 to 52.

In addition, the demand forecast display field 1310 displays a graph of the demand forecast before correction (dashed line graph), a graph of the demand forecast after correction (solid line graph), and an actual demand graph (dotted line graph). be done. Further, in the demand forecast display field 1310, a frame 1330 indicating a busy season is displayed.

The prediction basis display button 1320 is a button for displaying the correction basis screen 1400. When the user operates the prediction basis display button 1320, a correction basis screen 1400 is displayed.

The display unit 122 generates display information for displaying the correction basis screen 1400 based on the demand forecast information 350, the demand forecast information 340, and the busy season information 330.

The correction basis screen 1400 is a screen that presents the basis for correction of the demand forecast, and includes a demand forecast display field 1410, a peak distribution display field 1420, and a demand forecast (after correction) display field 1430.

The demand forecast display column 1410 is a column that displays a graph showing the demand forecast for the processing period of the target item. Demand forecast display column 1410 is displayed based on demand forecast information 340 and busy season information 330. The horizontal axis of the graph represents time-series identification information of prediction data during the processing period, and the vertical axis represents shipping amount. A dotted portion of the demand forecast display column 1410 indicates a busy season.

The peak distribution display column 1420 is a column that displays a graph showing the peak distribution of related items. The peak distribution display column 1420 is displayed based on the peak distribution information 321 and the busy season information 330. The horizontal axis of the graph represents time-series identification information of the demand time-series data during the processing period, and the vertical axis represents the number of peak occurrences. The dotted portion of the peak distribution display column 1420 indicates the busy season.

The demand forecast (after correction) display column 1430 is a column that displays a graph showing the demand forecast after correction of the processing period of the target item. The demand forecast (after correction) display column 1430 is displayed based on the demand forecast information 350 and the busy season information 330. The horizontal axis of the graph represents time-series identification information of prediction data during the processing period, and the vertical axis represents shipping amount. A dotted portion of the demand forecast (after correction) display field 1430 indicates a busy season. In addition, the demand forecast (after correction) display column 1430 displays the portion of the demand forecast that has been corrected and the reason for the correction.

According to the first embodiment, the demand forecasting system uses the peak distribution (demand trend) of related items to identify the busy season of the target item, and also uses the peak distribution (demand trend) of the related item to identify the busy season. Correct the predicted demand for the target item. This makes it possible to improve the accuracy of predicting the demand for the target item during the prediction period including the busy season.

Note that although the prediction unit 121 calculates the peak distribution using a plurality of data sets, it may also calculate the peak distribution using one data set. In this case, the peak distribution calculation process may be changed as follows.

In step S202, the peak distribution calculation unit 300 generates demand time series data from the shipping history group of one data set. In step S203, the peak distribution calculation unit 300 selects one demand time series data from among the plurality of demand time series data.

In step S204, the peak distribution calculation unit 300 identifies daily peaks. That is, the day when the daily shipment amount is larger than the demand threshold value is identified as the peak. The peak distribution calculation unit 300 stores aggregated data consisting of identification information of time series data and the number of peaks in a work area. That is, the number of peaks in one week is stored as aggregated data.

In step S205, the peak distribution calculation unit 300 determines whether processing of all demand time series data has been completed. If the processing of all demand time series data has not been completed, the peak distribution calculation unit 300 returns to step S203, and if the processing of all demand time series data has been completed, the peak distribution calculation unit 300 proceeds to step S206. . Similar processing is performed for related items.

The second embodiment differs from the first embodiment in the content of the demand forecasting process. The second embodiment will be described below, focusing on the differences from the first embodiment.

FIG. 15 is a diagram illustrating an example of the configuration of a demand forecasting system according to the second embodiment.

The demand forecasting system is composed of one computer 100. The hardware configuration of the computer 100 of the second embodiment is the same as that of the first embodiment. The configuration of the functional units included in the computer 100 of the second embodiment is the same as that of the first embodiment. The computer 100 of the second embodiment holds shipping history management information 130 and two pieces of predictive model management information 1500 and 1501. Note that the shipping history management information 130 is the same as in the first embodiment.

The prediction model management information 1500 stores definition information of a prediction model for predicting demand for items in a normal period. The prediction model is generated by a learning process using the shipping history of the normal period. Prediction model management information 1501 stores definition information of a prediction model for predicting demand for items during busy seasons. The model is generated through a learning process using shipping history during busy periods.

FIG. 16 is a diagram illustrating details of the functional units of the computer 100 according to the second embodiment.

The input section 120 of the second embodiment is the same as the input section 120 of the first embodiment.

The prediction unit 121 of the second embodiment includes a peak distribution calculation unit 300, a busy season calculation unit 301, and a demand prediction unit 302.

The peak distribution calculation section 300 and the busy season calculation section 301 of the second embodiment have the same functions as those of the first embodiment.

The demand prediction unit 302 predicts the demand for the target item based on the prediction model management information 1500 and 1501, the busy season information 330, and the shipping history information 310 of the target item, and outputs the prediction result as demand forecast information 340.

The display unit 122 presents information regarding the demand forecast for the target item based on the demand forecast information 340 output from the forecast unit 121.

The prediction unit 121 of the second embodiment differs from the first embodiment in that it does not perform prediction correction processing. The peak distribution calculation process of the second embodiment is the same as that of the first embodiment. The busy season calculation process of the second embodiment is the same as that of the first embodiment.

The demand forecasting process of the second embodiment is different from that of the first embodiment. FIG. 17 is a flowchart illustrating an example of the demand prediction process executed by the prediction unit 121 of the second embodiment.

The processing from step S401 to step S403 is the same as in the first embodiment.

The demand forecasting unit 302 determines whether a busy season exists (step S451). The process in step S451 is the same as the process in step S501.

If a busy season exists, the demand forecasting unit 302 acquires the forecast model management information 1500, 1501 (step S452).

The demand prediction unit 302 predicts the demand for the target item in the prediction period using the prediction model management information 1500, 1501 and the feature amount (step S453), and then proceeds to step S406.

Specifically, the demand forecasting unit 302 constructs a forecasting model defined in the forecasting model management information 1501, and inputs the feature amount calculated from the demand time series data of the busy season into the forecasting model. Further, the demand forecasting unit 302 constructs a forecasting model defined in the forecasting model management information 1500, and inputs the feature quantity calculated from the demand time series data of the normal period into the forecasting model.

If there is no busy season, the demand forecasting unit 302 acquires the forecast model management information 1500 (step S454).

The demand prediction unit 302 predicts the demand for the target item in the processing period using the prediction model management information 1500 and the feature amount (step S455), and then proceeds to step S406.

Specifically, the demand prediction unit 302 constructs a prediction model defined in the prediction model management information 1500, and inputs feature amounts to the prediction model.

The processing in step S406 is the same as in the first embodiment.

According to the second embodiment, the demand forecasting system identifies the busy season of the target item using the peak distribution (demand trend) of related items, and also switches the forecasting model for each of the busy season and normal period to Forecast the demand for a target item over a period of time. This makes it possible to improve the accuracy of predicting the demand for the target item during the prediction period including the busy season.

Note that the present invention is not limited to the above-described embodiments, and includes various modifications. Further, for example, the configurations of the embodiments described above are explained in detail in order to explain the present invention in an easy-to-understand manner, and the present invention is not necessarily limited to having all the configurations described. Further, a part of the configuration of each embodiment can be added to, deleted from, or replaced with other configurations.

Further, each of the above-mentioned configurations, functions, processing units, processing means, etc. may be partially or entirely realized in hardware by designing, for example, an integrated circuit. Further, the present invention can also be realized by software program codes that realize the functions of the embodiments. In this case, a storage medium on which a program code is recorded is provided to a computer, and a processor included in the computer reads the program code stored on the storage medium. In this case, the program code itself read from the storage medium realizes the functions of the embodiments described above, and the program code itself and the storage medium storing it constitute the present invention. Examples of storage media for supplying such program codes include flexible disks, CD-ROMs, DVD-ROMs, hard disks, SSDs (Solid State Drives), optical disks, magneto-optical disks, CD-Rs, magnetic tapes, A non-volatile memory card, ROM, etc. are used.

Further, the program code for realizing the functions described in this embodiment can be implemented in a wide range of program or script languages such as assembler, C/C++, Perl, Shell, PHP, Python, and Java (registered trademark).

Furthermore, by distributing the software program code that realizes the functions of the embodiment via a network, it can be stored in a storage means such as a computer's hard disk or memory, or a storage medium such as a CD-RW or CD-R. Alternatively, a processor included in the computer may read and execute the program code stored in the storage means or the storage medium.

In the above-described embodiments, the control lines and information lines are those considered necessary for explanation, and not all control lines and information lines are necessarily shown in the product. All configurations may be interconnected.

100 Computer 101 Input device 102 Output device 103 Network 110 CPU
111 IO interface 112 Network interface 113 Main storage device 114 Secondary storage device 120 Input section 121 Prediction section 122 Display section 130 Shipping history management information 131, 1500, 1501 Prediction model management information 300 Peak distribution calculation section 301 Busy season calculation section 302 Demand forecast Section 303 Correction section 310 Shipping history information 311 Shipping history information 320, 321 Peak distribution information 330 Busy season information 340, 350 Demand forecast information 350 Demand forecast information 1300 Demand forecast screen 1400 Correction basis screen

Claims (10)

A computer system comprising at least one calculator and predicting demand for an item in a prediction period,
a storage unit that stores history information that stores history regarding shipments per first unit period of each of the plurality of items, and definition information of a prediction model for predicting demand for the item;
Using the history of the first item and the history of a second item that has a relationship with the first item, a period that is part of the prediction period and in which the shipment amount of the first item is greater than a threshold value is determined. a busy season calculation unit that calculates a certain busy season;
Based on the definition information of the prediction model, the history of the first item, and the busy season, time-series data indicating the predicted shipment amount of the first item per second unit period of the prediction period is used as demand prediction information. a prediction unit that generates
a peak distribution calculation unit that calculates a peak distribution that is a distribution of the number of peaks in which the shipment amount of items per third unit period is larger than a threshold value in the prediction period;
Equipped with
The peak distribution calculation unit includes:
calculating a first peak distribution using the history of the first item;
calculating a second peak distribution using the history of the second item;
The busy season calculation section is
Calculating a mixed peak distribution based on the first peak distribution and the second peak distribution,
calculating the busy season based on the mixed peak distribution;
The prediction unit is
generating the demand forecasting information using the definition information of the forecasting model and the history of the first item;
updating the demand forecast information by correcting the time series data included in the busy season using the second peak distribution;
A computer system that outputs the updated demand forecast information . The computer system according to claim 1,
The computer system is characterized in that the busy season calculation unit refers to the mixed peak distribution and calculates, as the busy season, a continuous section in which the number of occurrences of the peak is greater than a busy season threshold. The computer system according to claim 2,
The computer system is characterized in that the busy season calculation unit calculates the busy season threshold based on statistical values calculated using the mixed peak distribution. The computer system according to claim 1,
A computer system comprising: a display unit that presents the demand forecast information before updating, the demand forecast information after update, and the busy season as a basis for correction. The computer system according to claim 1,
The storage unit stores definition information of a first prediction model for predicting demand for the item during the busy season, and definition information for a second prediction model for predicting demand for the item during the normal period. ,
The prediction unit is
Calculating the time series data of the busy season using the definition information of the first prediction model and the history of the first item,
Calculating the time series data for the normal period using the definition information of the second prediction model and the history of the first item,
A computer system that generates the time series data of the busy season and the time series data of the normal period as the demand forecast information. A method for predicting demand for items in a forecast period, the method being executed by a computer system including at least one computer, the method comprising:
The computer system manages history information that stores history regarding shipments per first unit period of each of a plurality of items, and definition information of a prediction model for predicting demand for the item,
The demand forecasting method for the above items is as follows:
The computer system,
Using the history of the first item and the history of a second item that has a relationship with the first item, a period that is part of the prediction period and in which the shipment amount of the first item is greater than a threshold value is determined. A first step of calculating a certain busy season;
Based on the definition information of the prediction model, the history of the first item, and the busy season, time-series data indicating the predicted shipment amount of the first item per second unit period of the prediction period is used as demand prediction information. a second step of generating as;
The first step is:
Using the history of the first item, calculating a first peak distribution that is a distribution of the number of peaks in which the shipment amount of the first item per third unit period is larger than a threshold value in the prediction period. Step 3 and
Using the history of the second item, calculate a second peak distribution that is a distribution of the number of peaks in which the shipment amount of the second item per third unit period is greater than a threshold value in the prediction period. The fourth step and
a fifth step of calculating a mixed peak distribution based on the first peak distribution and the second peak distribution;
a sixth step of calculating the busy season based on the mixed peak distribution,
The second step is
a seventh step of generating the demand forecast information using the definition information of the forecast model and the history of the first item;
an eighth step of updating the demand forecast information by correcting the time series data included in the busy season using the second peak distribution;
A method for predicting demand for an item, comprising: a ninth step of outputting the updated demand forecast information. A method for predicting demand for an item according to claim 6, comprising:
The sixth step refers to the mixed peak distribution and calculates, as the busy season, a continuous section in which the number of occurrences of peaks in which the shipment volume per third unit period is greater than the threshold is greater than the busy season threshold. A method for predicting demand for an item, comprising steps. A method for predicting demand for an item according to claim 7, comprising:
The method for predicting demand for an item, wherein the sixth step includes calculating the busy season threshold based on the statistical value calculated using the mixed peak distribution. A method for predicting demand for an item according to claim 6, comprising:
A method for predicting demand for an item, comprising the step of presenting the demand forecast information before updating, the demand forecast information after update, and the busy season as a basis for correction. A method for predicting demand for an item according to claim 6, comprising:
The computer system manages definition information of a first prediction model for predicting demand for the item during the busy season and definition information for a second prediction model for predicting demand for the item during the normal period. ,
The seventh step is
calculating the time series data of the busy season using the definition information of the first prediction model and the history of the first item;
calculating the time series data for the normal period using the definition information of the second prediction model and the history of the first item;
A method for predicting demand for an item, comprising the step of generating the time series data of the busy season and the time series data of the normal period as the demand forecast information.

Images