JP2012050241A - Demand predicting system and demand predicting method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a demand predicting system and a demand predicting method that can perform demand prediction while taking geographical correlation into consideration with a small calculation amount.SOLUTION: According to a demand predicting system 1, when a final demand prediction value (a second demand prediction value) for a prediction target area is to be determined, the correlation of actual demand values between the prediction target area and each of other areas is calculated, and an area having the correlation of a fixed value or more with the target area is extracted as a correlated area. The final demand prediction value (the second demand prediction value) of the prediction target area is calculated by using a regression equation (a second regression equation) obtained by regression analysis using the actual demand value of the correlated area and a first demand prediction value of the correlated area. As described above, the demand prediction of the prediction target area can be performed with a small calculation amount by using only the demand value of the area having high correlation with the prediction target area when geographical correlation is considered.

Description

  The present invention relates to a demand prediction system and a demand prediction method.

  Conventionally, various methods relating to demand prediction are known. For example, Patent Document 1 described below describes an apparatus that calculates a demand prediction value (specifically, a predicted sales number of tires) for each region. Specifically, this device calculates the predicted number of tires sold in a given period based on the predicted number of new tires, the estimated number of tires at the time of replacement, and the past tire sales rate. To do.

JP 2006-85645 A

  When obtaining a demand forecast value for each region as described in Patent Document 1, geographical correlation may be considered. However, such a geographical regression method requires a large amount of calculation and requires a long processing time.

  Therefore, an object of the present invention is to provide a demand prediction system and a demand prediction method capable of performing demand prediction in consideration of geographical correlation with a small amount of calculation.

  The demand prediction system according to the present invention includes, for each of a plurality of areas, a record value storage means for storing a past demand record value of the area, and a record of the demand record value of the area for each of the plurality of areas. First calculation means for generating a first regression equation by executing a regression analysis using the actual demand value, and calculating a first demand forecast value for the area using the first regression equation And for each of the plurality of areas, the predicted value storage means for storing the first demand predicted value calculated by the first calculation means, and the actual value storage means with one of the plurality of areas as the prediction target area An extraction means for calculating a correlation between the actual demand value of the prediction target area read from the demand actual value of another area and extracting the other area having the correlation equal to or greater than a predetermined threshold as a correlation area; In A generation unit that generates a second regression equation by executing a regression analysis using the actual demand value of the correlation area extracted, a second regression equation generated by the generation unit, and a read from the predicted value storage unit Second calculation means for calculating the second demand prediction value of the prediction target area using the demand prediction value of the correlation area, and output means for outputting the second demand prediction value extracted by the second calculation means And comprising.

  The demand prediction method of the present invention is a demand prediction method executed by a demand prediction system, and for each of a plurality of areas, an actual value storage step of storing past demand actual values of the area in an actual value storage means; For each of the plurality of areas, the actual demand value of the area is read from the actual value storage means, a regression analysis is performed using the actual demand value, and a first regression equation is generated, and the first regression equation is generated. The first calculation step of calculating the first demand prediction value of the area using the first and the first demand prediction value calculated in the first calculation step for each of the plurality of areas is stored in the prediction value storage means. A predicted value storage step, and a demand actual value of the prediction target area read from the actual value storage means and a demand actual value of another area as one of a plurality of areas as a prediction target area A second step of calculating a function, extracting another area having the correlation equal to or greater than a predetermined threshold as a correlation area, and performing a regression analysis using the actual demand value of the correlation area extracted in the extraction step. The second demand prediction of the prediction target area using the generation step of generating the regression equation, the second regression equation generated in the generation step, and the demand prediction value of the correlation area read from the prediction value storage means A second calculation step of calculating a value; and an output step of outputting the second demand prediction value extracted in the second calculation step.

  According to such an invention, when the final demand forecast value (second demand forecast value) of a certain forecast target area is obtained, the demand actual value between the forecast target area and other areas is calculated. A correlation is calculated, and an area where the correlation is a certain level or more is extracted as a correlation area. Then, using the regression equation (second regression equation) obtained by the regression analysis using the actual demand value of the correlation area and the first demand prediction value of the correlation area, the final prediction target area is obtained. A demand forecast value (second demand forecast value) is calculated. In this way, when considering geographical correlation, it is possible to perform demand prediction of the prediction target area with a small amount of calculation by using only the demand values of other areas having high correlation with the prediction target area. Become.

  In the demand prediction system of the present invention, the extracting means extracts another area whose distance from the prediction target area is a predetermined value or less as a peripheral area, the demand actual value of the prediction target area and the demand actual value of the peripheral area, And the surrounding area where the correlation is equal to or greater than a threshold may be extracted as a correlation area.

  In this case, after the area that is a candidate for the correlation area is narrowed down in advance to be located near the prediction target area, the correlation with the prediction target area is calculated only for the narrowed candidates. Thereby, the amount of calculating the correlation can be reduced.

  According to such a demand forecasting system and demand forecasting method, the demand forecast value of the forecast target area is calculated using only the demand value of the other area having a high correlation with the forecast target area. It is possible to carry out demand forecast in consideration with a small amount of calculation.

It is a figure which shows the function structure of the demand prediction system which concerns on embodiment. It is a figure which shows the hardware constitutions of the demand prediction system shown in FIG. It is a flowchart which shows the production | generation method of the data for analysis. It is a flowchart which shows the production | generation method of 1st regression type data. It is a flowchart which shows the production | generation method of the data for prediction. It is a flowchart which shows the production | generation method of 1st prediction result data. It is a flowchart which shows the extraction method of a correlation area. It is a flowchart which shows the production | generation method of 2nd regression type data. It is a flowchart which shows the production | generation method of 2nd prediction result data. It is a flowchart which shows operation | movement of the demand prediction system shown in FIG.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the description of the drawings, the same or equivalent elements are denoted by the same reference numerals, and redundant description is omitted.

  First, the function and configuration of the demand prediction system 1 according to the embodiment will be described with reference to FIGS. The demand prediction system 1 is a computer system that predicts demand in each of a plurality of areas. In particular, in the present embodiment, the demand prediction system 1 will be described as predicting taxi demand (the objective variable is the number of taxi passengers).

  As shown in FIG. 1, the demand prediction system 1 includes, as functional components, an analysis data generation unit 11, an analysis data storage unit (actual value storage unit) 12, and a first regression equation generation unit (first calculation unit) 13. , Prediction data generation unit (first calculation unit) 14, prediction data storage unit 15, first prediction result generation unit (first calculation unit) 16, first prediction result storage unit (prediction value storage unit) 17, area An extraction unit (extraction unit) 18, a second regression equation generation unit (generation unit) 19, a second prediction result generation unit (second calculation unit, output unit) 20, and a second prediction result storage unit 21 are provided.

  The demand prediction system 1 may be configured with a single server or a plurality of servers (distributed systems). In any case, the server V used in the demand prediction system 1 has a hardware configuration as shown in FIG. That is, the server V includes a CPU 101 that executes an operating system, application programs, and the like, a main storage unit 102 that includes a ROM and a RAM, an auxiliary storage unit 103 that includes a hard disk, a network card, and the like. The communication control unit 104 includes an input unit 105 such as a keyboard and a mouse, and an output unit 106 such as a monitor.

  Each functional component of the demand prediction system 1 shown below reads predetermined software on the CPU 101 or the main storage unit 102, and under the control of the CPU 101, the communication control unit 104, the input unit 105, the output unit 106, and the like. This is realized by reading and writing data in the main storage unit 102 and the auxiliary storage unit 103. Data and databases necessary for processing are stored in the main storage unit 102 and the auxiliary storage unit 103.

  Returning to FIG. 1, the analysis data generation unit 11 is means for generating analysis data in the procedure as shown in FIG. 3. That is, the analysis data generation unit 11 first generates mesh ride data in which the actual number of rides (actual value of the number of taxi passengers) is tabulated for each area within a predetermined range based on the mesh data D11 and the ride data D12. (Step S11). Subsequently, the analysis data generation unit 11 generates the analysis data D17 based on the mesh boarding data, the population distribution data D13, the weather data D14, the temperature data D15, and the time zone data D16 (step S12).

  The mesh data D11 is data relating to the area, and is stored in the demand prediction system 1 in advance. Each set area is, for example, a 500M square (mesh), but the shape and size of the area are not limited to this. For example, each area may be a 100M square, or an area having an arbitrary shape other than a rectangle may be set. In addition, the shape and size of all the areas need not be uniform. For example, an area suitable for each region may be set based on the population density or the geographical situation of the region. In any case, a plurality of mesh data D11 relating to a plurality of areas are prepared.

  The mesh data D11 is a record in which an area ID that identifies an area, an area polygon that indicates the shape and size of the area, and a center point that indicates the latitude and longitude of the center point of the area are associated with each other.

  The boarding data D12 is a record formed by associating points and times when passengers get on the taxi. In addition, the collection method of boarding data D12 is not limited. For example, the position and time of the boarding may be estimated based on the change in the position of the passenger's mobile phone obtained from the mobile communication network, and recorded as boarding data, or static information obtained through questionnaire surveys, etc. May be converted into data as boarding data.

  The analysis data generation unit 11 generates mesh ride data by counting the number of rides for each area and time from the mesh data D11 and the ride data D12. This mesh boarding data is a record in which an area ID, area polygon, center point, time, and actual boarding number are associated with each other. The unit time for counting the number of rides may be arbitrarily determined. For example, the total unit time may be set to 5 minutes, 1 hour, 1 day, and the like.

  Subsequently, the analysis data generation unit 11 generates analysis data D17. Here, each data used at this time will be described.

  The population distribution data D13 is data in which an area ID, an area polygon, time, and a population in the area at that time are associated with each other. The weather data D14 is data in which the area ID, area polygon, time, and precipitation in the area at that time are associated with each other. The temperature data D15 is data in which an area ID, an area polygon, a time, and the temperature in the area at that time are associated with each other. The time zone data D16 is data in which time, day of the week, and weekday or holiday classification are associated with each other. In addition, the collection method of these four types of data is not limited, For example, the demand prediction system 1 may receive each data from the other system which provides these data, and may memorize | store it beforehand. The time in each data is set in the same manner as the time in the mesh ride data.

  Based on the generated mesh ride data and the population distribution data D13, weather data D14, temperature data D15, and time zone data D16, the analysis data generation unit 11 is configured to calculate the number of passengers, population, Analytical data D17 indicating the amount and temperature is generated. That is, the analysis data D17 is data in which the area ID, area polygon, center point, time, number of passengers, population, precipitation, and temperature are associated with each other. The number of rides in the analysis data D17 can be said to be a past demand actual value. The analysis data generation unit 11 stores the generated analysis data D17 in the analysis data storage unit 12.

  The data used when generating the analysis data D17 is not limited to the population, precipitation, temperature, and day of the week as described above, and other data may be used. For example, analysis data may be generated using other factors such as stock prices.

  The analysis data storage unit 12 is means for storing the analysis data D17. Since the analysis data generation unit 11 generates analysis data at each time in each area, the analysis data storage unit 12 stores the actual value of the number of passengers per hour for each of the plurality of areas. Become.

  The first regression equation generation unit 13 is means for determining a variable (explanatory variable) to be used for regression analysis in the procedure as shown in FIG. 4 and generating a first regression equation using the variable.

  That is, the first regression equation generation unit 13 calculates the total number of passengers and the average values of population, precipitation, and temperature for each time for the analysis data D17 stored in the analysis data storage unit 12. Is obtained (step S21). Subsequently, the first regression equation generation unit 13 selects some variables from the analysis data D17 (step S22), and evaluates the result of the regression analysis using the selected variables by cross-validation (step S23). The first regression equation generation unit 13 executes the processing of steps S22 and S23 for various combinations of variables, and finally determines the combination of variables that has obtained the highest evaluation (step S24).

  Subsequently, the first regression equation generation unit 13 performs the regression analysis on the analysis data D17 using the determined combination of variables to generate the first regression equation data D18 (step S25). The first regression equation data D18 is data obtained by associating the area ID, the area polygon, the center point, and the obtained first regression equation. When the first regression equation data relating to each area is generated, the first regression equation generation unit 13 outputs the first regression equation data to the first prediction result generation unit 16.

  Although multiple regression analysis is used in the present embodiment, the specific method of regression analysis is not limited. For example, a neural network or a method described in Japanese Patent Application No. 2009-265512 is described. May be used.

  The prediction data generation unit 14 is means for generating prediction data in the procedure as shown in FIG. That is, the prediction data generation unit 14 generates the prediction data D22 from the mesh data D11, the population distribution prediction data D19, the weather prediction data D20, the temperature prediction data D21, and the time zone data D16 (step S31). The generation of the prediction data D22 is similar to the generation of the analysis data D17. However, the analysis data D17 is generated using past data, that is, a known actual value, whereas the prediction data D22 is generated in the future. Therefore, it is generated using forecast data and forecast data.

  The data structures of the population distribution prediction data D19, the weather prediction data D20, and the temperature prediction data D21 are the same as the population distribution data D13, the weather data D14, and the temperature data D15, respectively. The value is a predicted value. The prediction data D22 is data in which an area ID, an area polygon, a center point, time, a predicted population, an estimated precipitation amount, and an estimated temperature are associated with each other.

  When the prediction data at each time in each area is generated, the prediction data generation unit 14 stores these prediction data in the prediction data storage unit 15.

  The prediction data storage unit 15 is means for storing the prediction data D22. The prediction data D22 stored in the prediction data storage unit 15 is referred to by a first prediction result generation unit 16 and a second prediction result generation unit 20 described later.

  The 1st prediction result production | generation part 16 is a means which calculates the 1st prediction boarding number (1st demand prediction value) of each area using a 1st regression equation in the procedure as shown in FIG. That is, the first prediction result generation unit 16 applies the prediction data D22 of the area to the first regression equation indicated by the first regression equation data D18 of the one area, so that the first prediction result of the area Prediction result data D23 is generated (step S41), and the result data D23 is stored in the first prediction result storage unit 17. The 1st prediction result production | generation part 16 performs such a process about each area. The first prediction result data D23 is data obtained by associating the area ID, area polygon, center point, time, and first predicted number of rides (first demand predicted value). The 1st prediction result production | generation part 16 calculates | requires the predicted value of the boarding number in each time of each area, and stores it in the 1st prediction result memory | storage part 17 as 1st prediction result data D23.

  The first prediction result storage unit 17 is means for storing the first prediction result data D23. The 1st prediction result storage part 17 memorizes the 1st prediction boarding number for every time about each of a plurality of areas. The predicted number of passengers stored in the first prediction result storage unit 17 is an intermediate result derived in the process of obtaining the second predicted number of passengers (second predicted demand value) that is the final result.

  The area extracting unit 18 is means for extracting another area having a high correlation with the prediction target area in the procedure as shown in FIG. First, the area extraction unit 18 sets one of a plurality of areas as a prediction target area. Subsequently, the area extraction unit 18 refers to the mesh data D11 and extracts another area whose distance from the prediction target area is a predetermined value or less. Hereinafter, the extracted area is referred to as a “peripheral area”. The reason for extracting the surrounding area is that an area that has a high correlation with a certain area often exists in the vicinity of the one area, and that the calculation amount increases when the correlation with all the areas is obtained. This is because the objects for which correlation is to be obtained are narrowed down in advance.

  In addition, as a method for obtaining the distance between areas, for example, a method of setting a distance between center point points as a distance between areas, or a minimum distance among the boundaries between areas is set as a distance between areas. Although the method etc. to employ | adopt are mentioned, it is not limited to these.

ここで、ｘ_ｉは予測対象エリアの実績乗車数であり、ｙ_ｉは一の周辺エリアの実績乗車数である。また、

は予測対象エリアの全時間における実績乗車数の相加平均であり、

は一の周辺エリアの全時間における実績乗車数の相加平均である。 Subsequently, the area extraction unit 18 reads the analysis data D17a for the prediction target area and the analysis data D17b for the peripheral area from the analysis data storage unit 12, and the correlation of the actual number of rides between the prediction target area and the peripheral area. The number is obtained by the following equation (1) (step S51). The range of the correlation coefficient is −1 to 1.

Here, x _i is the actual number of rides in the prediction target area, and y _i is the actual number of rides in one peripheral area. Also,

Is the arithmetic average of the actual number of rides over the entire forecast area,

Is the arithmetic average of the actual number of passengers in the entire surrounding area.

  Subsequently, the area extraction unit 18 determines whether or not the absolute value of the obtained correlation coefficient is equal to or greater than a predetermined threshold (step S52), and extracts a peripheral area whose absolute value is equal to or greater than the threshold as a correlation area. (Step S53). An arbitrary value may be used as the threshold value.

  The area extraction unit 18 performs such processing for all peripheral areas (step S54), and calculates the area ID (prediction target area ID) of the prediction target area and the area ID (correlation area ID) of one or more correlation areas. Correlation area data D24 that is a combination is generated. The area extraction unit 18 generates such correlation area data D24 for each of the areas defined by the mesh data D11 and outputs the correlation area data D24 to the second regression equation generation unit 19.

  The second regression equation generation unit 19 executes the regression analysis using the explanatory variables used in the first regression equation generation unit 13 and the actual number of rides in the correlation area in the procedure as shown in FIG. A means for generating a regression equation. That is, the second regression equation generation unit 19 newly generates a regression equation after adding the actual number of rides in the correlation area to the explanatory variables.

  The second regression equation generation unit 19 performs the following processing for each prediction target area. That is, the second regression equation generation unit 19 reads out the analysis data D17a corresponding to the prediction target area ID indicated by the correlation area data from the analysis data storage unit 12. In addition, the second regression equation generation unit 19 reads analysis data corresponding to the correlation area ID indicated by the correlation area data from the analysis data storage unit 12 and acquires the actual number of rides in one or more correlation areas. Subsequently, the second regression equation generation unit 19 performs a regression analysis on the analysis data D17a in the prediction target area using the explanatory variables used in the first regression equation generation unit 13 and the read actual number of rides as the current explanatory variables. Second regression equation data D25 is generated (step S61). In the second regression equation data D25, the area ID, area polygon and center point of the prediction target area, the obtained second regression equation, and one or more correlation area IDs corresponding to the prediction target area are mutually connected. It is data that is related.

  The second regression equation generation unit 19 executes such processing for all the prediction target areas, and outputs the second regression equation data D25 of each prediction target area to the second prediction result generation unit 20.

  The second prediction result generation unit 20 uses the second regression equation and the first predicted number of rides in the correlation area in the procedure as shown in FIG. 9 to obtain the final second predicted ride in the prediction target area. It is a means for calculating the number.

  The second prediction result generation unit 20 performs the following process for each prediction target area. That is, the second prediction result generation unit 20 reads the prediction data D22a corresponding to the area ID of the prediction target area from the prediction data storage unit 15. In addition, the second prediction result generation unit 20 reads the first prediction result data D23a corresponding to the correlation area ID associated with the prediction target area from the first prediction result storage unit 17. Subsequently, the second prediction result generation unit 20 applies the read prediction data D22a and the first prediction result data D23a to the second regression equation indicated by the second regression equation data D25a of the prediction target area. Then, the second prediction result data D26 is generated (step S71). Then, the second prediction result generation unit 20 stores the second prediction result data D26 in the second prediction result storage unit 21. The second prediction result data D26 is data in which the area ID, area polygon, center point, time, and second predicted number of rides (second demand predicted value) of the prediction target area are associated with each other.

  The second prediction result generation unit 20 performs such processing for all the prediction target areas, and stores the second prediction result data D26 of each prediction target area in the second prediction result storage unit 21.

  The second prediction result storage unit 21 is means for storing second prediction result data D26. The 2nd prediction result storage part 21 memorizes the 2nd prediction boarding number for every time about each of a plurality of areas. The predicted number of passengers stored in the second prediction result storage unit 21 is the final prediction result.

  Next, the operation of the demand prediction system 1 shown in FIG. 1 will be described using FIG. 10 and the demand prediction method according to the present embodiment will be described. The demand prediction system 1 determines the demand prediction value of each area by performing a two-stage prediction process.

  In the first-stage prediction process, first, the analysis data generation unit 11 generates analysis data for each area and stores it in the analysis data storage unit 12 (step S81, actual value storage step). Then, the 1st regression type production | generation part 13 produces | generates a 1st regression formula about each area by performing the regression analysis with respect to the data for an analysis of the said area (step S82, 1st calculation step). Subsequently, the prediction data generation unit 14 generates prediction data for each area (step S83, first calculation step). Subsequently, the first prediction result generation unit 16 calculates the first predicted number of rides by applying the analysis data to the first regression equation for each area (step S84, first calculation step), The first prediction result data is stored in the first prediction result storage unit 17 (prediction value storage step).

  Next, a second stage prediction process is performed. First, the area extraction unit 18 extracts a correlation area for each prediction target area (step S85, extraction step). This correlation area is a peripheral area whose correlation with the prediction target area regarding the actual number of rides is a predetermined value or more. Subsequently, the second regression equation generation unit 19 performs, for each prediction target area, a regression analysis in which the actual number of rides in the correlation area is added as an explanatory variable to the analysis data of the prediction target area. A regression equation is generated (step S86, generation step). Then, the second prediction result generation unit 20 applies, for each prediction target area, the analysis data of the prediction target area and the first prediction result data of the correlation area to the second regression equation. The predicted number of passengers is calculated (step S87, second calculation step). The second prediction result generation unit 20 stores the calculated second predicted number of rides in the second prediction result storage unit 21 (output step). Thereby, the estimated number of boarding of each area is decided.

  As described above, according to the present embodiment, when obtaining the final predicted number of passengers in a certain prediction target area, the correlation of the actual number of passengers is calculated between the prediction target area and the surrounding area. , Areas whose correlation is above a certain level are extracted as correlation areas. Then, using the second regression equation obtained by the regression analysis using the actual demand value of the correlation area and the first predicted number of rides of the correlation area, the final second of the prediction target area is obtained. The predicted number of rides is calculated. In this way, when considering geographical correlation, it is possible to perform demand prediction of the prediction target area with a small amount of calculation by using only the demand values of other areas having high correlation with the prediction target area. Become.

  Further, according to the present embodiment, the correlation area candidates are narrowed down in advance to those that are close to the prediction target area, and only the narrowed candidates (peripheral areas) are correlated with the prediction target area. Is calculated. Thereby, the amount of calculating the correlation can be reduced.

  The present invention has been described in detail based on the embodiments. However, the present invention is not limited to the above embodiment. The present invention can be variously modified without departing from the gist thereof.

  In the above embodiment, the demand prediction system 1 predicts the number of taxi passengers. However, the type of demand to be predicted is not limited to this, and the present invention can be applied to various demand predictions. For example, the present invention can be applied to prediction of sales of products (the objective variable is the number or amount of sales) and future population prediction (the objective variable is population).

  In the above embodiment, the area extraction unit 18 first extracts a peripheral area for one prediction target area, and extracts a correlation area by obtaining a correlation of the actual number of rides with the prediction target area for each peripheral area. However, the method of extracting the correlation area is not limited to this. For example, the extraction means obtains the correlation of the actual number of rides with all other areas for one prediction target area without extracting the surrounding area, and extracts the correlation area from all the other areas. Also good. That is, the extracting means may extract the correlation area based only on the correlation of the actual number of rides without considering the distance between the areas.

  In the above embodiment, the second predicted number of rides that is the final result is stored in the second prediction result storage unit 21, but the method of outputting the second demand predicted value is not limited to this. For example, the second demand prediction value may be transmitted to another information processing system or a terminal.

  DESCRIPTION OF SYMBOLS 1 ... Demand prediction system, 11 ... Analytical data generation part, 12 ... Analytical data storage part (actual value storage means), 13 ... 1st regression type production | generation part (1st calculation means), 14 ... Prediction data generation part (First calculation means), 15 ... prediction data storage section, 16 ... first prediction result generation section (first calculation means), 17 ... first prediction result storage section (prediction value storage means), 18 ... area extraction section (Extraction means), 19... Second regression equation generation section (generation means), 20... Second prediction result generation section (second calculation means, output means), 21.

Claims (3)

For each of the plurality of areas, actual value storage means for storing the past demand actual value of the area;
For each of the plurality of areas, the actual demand value of the area is read from the actual value storage means, a regression analysis using the actual demand value is performed, a first regression equation is generated, and the first regression First calculating means for calculating a first demand forecast value of the area using an equation;
Predicted value storage means for storing the first demand predicted value calculated by the first calculation means for each of the plurality of areas;
Using one of the plurality of areas as a prediction target area, a correlation between the demand actual value of the prediction target area read from the actual value storage unit and the demand actual value of another area is calculated, and the correlation is predetermined. Extracting means for extracting the other area that is equal to or greater than the threshold value as a correlation area;
Generating means for generating a second regression equation by performing regression analysis using the actual demand value of the correlation area extracted by the extracting means;
A second demand calculation value for the prediction target area is calculated using the second regression equation generated by the generation means and the demand prediction value for the correlation area read from the prediction value storage means. A calculation means;
Output means for outputting a second demand forecast value extracted by the second calculation means;
A demand forecasting system comprising: The extraction means extracts another area whose distance from the prediction target area is a predetermined value or less as a peripheral area, calculates a correlation between the demand actual value of the prediction target area and the demand actual value of the peripheral area, Extracting the peripheral area where the correlation is equal to or greater than the threshold as the correlation area;
The demand forecasting system according to claim 1 characterized by things. A demand forecast method executed by a demand forecast system,
For each of the plurality of areas, an actual value storage step of storing the past demand actual value of the area in the actual value storage means;
For each of the plurality of areas, the actual demand value of the area is read from the actual value storage means, a regression analysis using the actual demand value is performed, a first regression equation is generated, and the first regression A first calculation step of calculating a first demand forecast value of the area using an equation;
For each of the plurality of areas, a predicted value storage step of storing the first demand predicted value calculated in the first calculation step in a predicted value storage means;
Using one of the plurality of areas as a prediction target area, a correlation between the demand actual value of the prediction target area read from the actual value storage unit and the demand actual value of another area is calculated, and the correlation is predetermined. An extraction step of extracting the other area that is equal to or greater than the threshold value as a correlation area;
A generation step of generating a second regression equation by performing a regression analysis using the actual demand value of the correlation area extracted in the extraction step;
A second demand calculation value for the prediction target area is calculated using the second regression equation generated in the generation step and the demand prediction value for the correlation area read from the prediction value storage means. A calculation step;
An output step of outputting the second demand forecast value extracted in the second calculation step;
Demand forecasting method characterized by including.

Images