JP2007018216A - Reasonable inventory quantity calculation system and reasonable inventory quantity calculation method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To calculate reasonable inventory quantity under the consideration of the temporal change of merchandise characteristics based on the past time-sequential data. <P>SOLUTION: This reasonable inventory quantity calculation system comprises: steps S411 to S413 for computing a regression formula with sales quantity data as criterion variables and with factor data as explanatory variables with respect to each of a plurality of predetermined object periods; a step S415 for selecting the regression formula for calculating a sales quantity decision term prediction value as the prediction value of sales quantity in the optimal regression formula decision term from the factor data in the optimal regression formula decision term by each regression formula, and calculating the sales quantity decision term prediction value which is the closest to the value in the optimal regression formula decision term of the sales quantity data as the optimal regression formula; and a step S416 for calculating the sales quantity pertinent term prediction value as a prediction value in the current term of the sales quantity by the optimal regression formula from the current factor data. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an appropriate inventory quantity calculation system and an appropriate inventory calculation method, and more particularly to an appropriate inventory quantity calculation system and an appropriate inventory quantity calculation method for calculating an appropriate inventory quantity based on sales results and factors that affect sales.

  Having the right inventory is an important issue for business management because excess inventory leads to losses and missing items lead to lost profits. In particular, it is important to accurately perform sales forecasts that are the basis for determining the appropriate inventory amount. Here, if sales prediction is left to the experience and intuition of the person in charge, there is a problem that the success or failure of the prediction depends on the person in charge and it is difficult to inherit skills. On the other hand, in order to predict sales based on data, sales are usually affected by multiple factors, and the number of products handled in recent years has increased, so automation is attempted using computers. It is essential. For this reason, various appropriate inventory quantity calculation systems or sales prediction systems have been conventionally known.

  Patent Document 1 discloses a sales volume prediction device that predicts the sales volume of a product to be sold from quantitative values such as past sales performance quantities.

  Moreover, in patent document 2, it is provided with the calculation means and the memory | storage means, and based on the past sales data memorize | stored in the memory | storage means, the past weather data, and the past external factor data, A forecasting system for generating sales figures according to weather conditions and external factor conditions is disclosed.

  Further, in Patent Document 3, a data series of dependent variables and a data series of explanatory variables are set from information on past shipment volume results and weather conditions, and a multiple regression model adapted to the dependent variable data series and the explanatory variable data series is set. A system that generates and outputs a predicted shipment amount is disclosed.

  Furthermore, in Patent Document 4, a data series is created by processing information such as weather, date, day of the week, time, etc., and all the created data is subjected to multiple regression analysis to select information with a high degree of contribution to prediction. A merchandise sales forecasting device that forecasts merchandise sales based on the information thus disclosed is disclosed.

  Furthermore, in Patent Document 5, the previous week's week index that compares the sales performance of the previous week corresponding to the forecast week that is the target of sales forecast and the sales performance of the corresponding week of the previous week is calculated, Calculate a correction value that compares the weighted average over the most recent period of the forecast year with the corresponding weekly average of the previous year, and then multiply the previous week index by the correction value to calculate the forecast index, An automatic product ordering system that predicts the order quantity from a forecast index and the actual value of each day of the week is disclosed.

Japanese Patent Laid-Open No. 7-175786 JP 2004-272673 A JP 2005-78277 A JP-A-8-212191 JP-A-9-198448

  An object of the present invention is an appropriate inventory quantity calculation system and an appropriate inventory calculation method for calculating an appropriate inventory quantity of a product based on past time-series data, and calculates an appropriate inventory quantity in consideration of a time change of product characteristics. It is to provide an appropriate inventory quantity calculation system and an appropriate inventory quantity calculation method.

  Another object of the present invention is to provide an appropriate inventory quantity calculation system and an appropriate inventory quantity calculation method for calculating an appropriate inventory quantity by selecting an optimal sales quantity prediction method for each product.

  Still another object of the present invention is to provide an appropriate inventory quantity calculation system and an appropriate inventory quantity calculation method capable of calculating an appropriate inventory quantity with high prediction accuracy even for a product whose sales volume varies greatly with time. It is.

  Hereinafter, means for solving the problem will be described using the numbers used in (Best Mode for Carrying Out the Invention). These numbers are added to clarify the correspondence between the description of (Claims) and (Best Mode for Carrying Out the Invention). However, these numbers should not be used to interpret the technical scope of the invention described in (Claims).

  The appropriate inventory quantity calculation system of the present invention includes an arithmetic device (31), a storage device (33), input devices (4, 5), and an output device (12). Here, the storage device stores sales volume data that is time-series data of sales volume before the previous period and factor data that is time-series numerical data of the factors affecting the sales volume before and after the previous period. A database (306), a sales performance database (302) that stores sales volume data source data and factor data source data, both of which are time series data, and a current period that is numerical data of predicted values or scheduled values of the factors for the current period The factor database (301, 303) for storing factor data is stored. And the said arithmetic unit stores the said sales quantity data original data and the said factor data original data which are input from the said input device in the said sales performance database. Further, the arithmetic device reads the sales volume data source data and the factor data source data from the sales performance database, and aggregates the sales volume data source data and the factor data source data for each period to calculate the sales. Quantity data and the factor data are created respectively. Then, the generated sales quantity data and the factor data are stored in the time series database. The arithmetic device stores the current period factor data input from the input device in the current period factor database. The arithmetic device reads the sales quantity data and the factor data from the time-series database, uses the sales quantity data as an objective variable for each of a plurality of predetermined target periods, and uses the factor data as an explanatory variable. The regression equation is calculated. Further, the arithmetic device reads the current factor data from the current factor database, and determines the optimal regression equation judgment period based on each of the regression equations from the factor data in the optimal regression equation judgment period which is a predetermined period before the previous period. A sales volume judgment period prediction value, which is a predicted value of the sales volume, is calculated. Further, the arithmetic unit selects, as the optimal regression formula, the regression formula that calculates the sales quantity judgment period predicted value closest to the value in the optimal regression formula judgment period of the sales quantity data. Further, the arithmetic unit calculates a sales volume current forecast value, which is a forecast value of the sales volume in the current period, from the current factor data by the optimal regression equation. Further, the arithmetic unit calculates a proper inventory quantity by adding a predetermined safety inventory quantity to the sales volume current period forecast value. The output device outputs the appropriate inventory amount.

  In the appropriate inventory amount calculation system according to the present invention, each of the plurality of predetermined target periods may have different start periods and the same end periods.

  In the appropriate inventory quantity calculation system of the present invention, the storage device stores a correlation coefficient / rank correspondence table (304) that defines the correspondence between the correlation coefficient and the rank of the degree of correlation. A plurality of the factors exist, and the factor data exists corresponding to each of the factors. The arithmetic device calculates a correlation coefficient between the sales quantity data and each of the factor data for the longest target period having the longest period among the plurality of predetermined target periods. Further, the arithmetic device ranks each of the factor data from the correlation coefficient with reference to the correlation coefficient / rank correspondence table. When the ranking satisfies a predetermined first condition, the calculation device calculates the regression equation, which is a multiple regression equation, using the factor data with a rank higher than a predetermined rank as the explanatory variable.

  In the appropriate inventory amount calculation system according to the present invention, when the ranking satisfies a predetermined second condition, the arithmetic unit uses the factor data assigned the highest rank as the explanatory variable with a single regression equation. A certain regression equation is calculated.

  In the appropriate inventory quantity calculation system of the present invention, the storage device stores a first standard deviation / smoothing constant correspondence table (305) that defines the correspondence between the first standard deviation and the smoothing constant. In the case where the ranking does not satisfy both the first condition and the second condition, the computing device determines the sales volume number that is the sales volume data included in the predetermined first target period that ends in the previous period. The average value of 1 target period data is calculated. Further, the arithmetic unit calculates the sales volume increase / decrease degree data by dividing each of the sales volume first target period data by the sales volume data one period before the sales volume first target period data. Further, the arithmetic unit calculates the first standard deviation which is a standard deviation of the sales volume increase / decrease degree data. Furthermore, the arithmetic unit determines the smoothing constant from the first standard deviation with reference to the first standard deviation / smoothing constant correspondence table. Furthermore, instead of calculating the sales volume current period forecast value by the optimal regression equation, the arithmetic unit uses an exponential smoothing formula from the determined smoothing constant, the average value, and the sales volume data in the previous period. The sales volume current period forecast value is calculated.

  In the appropriate inventory quantity calculation system according to the present invention, the arithmetic unit calculates the sales quantity based on the optimal regression equation from the factor data included in the second target period, which is the target period in which the optimal regression equation is calculated. Time series data of the sales quantity second target period predicted value, which is the predicted value in the second target period, is calculated. Further, the arithmetic device takes the difference between the time series data of the sales volume second target period predicted value and the time series data of the sales quantity data in the second target period, thereby predicting the prediction by the optimal regression equation. Time series data of the regression equation prediction error which is an error is calculated. Further, the arithmetic unit calculates a second standard deviation which is a standard deviation of time series data of the regression equation prediction error, and calculates the safety stock quantity by multiplying the second standard deviation by a predetermined constant.

  In the appropriate inventory amount calculation system according to the present invention, the arithmetic unit calculates the first target period of the sales quantity from the smoothing constant, the average value, and the sales quantity first target period data by the exponential smoothing formula. The time series data of the sales quantity first target period prediction value that is the prediction value at is calculated. Further, the arithmetic device takes the difference between the time series data of the sales volume first target period predicted value and the time series data of the sales quantity data in the first target period to perform prediction by the exponential smoothing equation. Time series data of exponential smoothing type prediction error which is a prediction error is calculated. Further, the arithmetic unit calculates a third standard deviation that is a standard deviation of time series data of the exponential smoothing type prediction error, and calculates the safety stock quantity by multiplying the third standard deviation by a predetermined constant. .

  The appropriate inventory quantity calculation method of the present invention is executed by an appropriate inventory quantity calculation system including an arithmetic device (31), a storage device (33), an input device (4, 5), and an output device (12). This is an appropriate inventory amount calculation method. Here, the storage device stores the sales volume data which is the previous time series data of the sales volume and the factor data which is the previous time series numerical data of the factor affecting the sales volume. (306), a sales performance database (302) that stores sales volume data source data and factor data source data, both of which are time series data, and current factor that is numerical data of predicted values or scheduled values of the factors in the current term The current period factor database (301, 303) for storing data is stored. And the appropriate inventory quantity calculation method of this invention includes the said arithmetic unit storing the said sales volume data original data and the said factor data original data which are input from the said input device in the said sales performance database. . Further, according to the method for calculating an appropriate inventory amount of the present invention, the arithmetic device reads the sales quantity data source data and the factor data source data from the sales performance database, and the sales quantity data source data and the factor data source data. And the sales volume data and the factor data are respectively created. In the method for calculating the appropriate inventory amount according to the present invention, the arithmetic device stores the created sales quantity data and the factor data in the time series database. In addition, the appropriate inventory amount calculation method of the present invention includes that the arithmetic device stores the current period factor data input from the input device in the current period factor database. Then, in the appropriate inventory quantity calculation method of the present invention, the arithmetic device reads the sales quantity data and the factor data from the time-series database, and the sales quantity data for each of a plurality of predetermined target periods. It includes calculating a regression equation that uses the factor data as an explanatory variable as an objective variable. Furthermore, in the method for calculating the appropriate inventory amount according to the present invention, the arithmetic unit reads the current factor data from the current factor database, and the regression equation is calculated from the factor data in an optimal regression equation determination period that is a predetermined period before the previous period. Respectively, calculating a sales quantity judgment period prediction value that is a prediction value of the sales quantity in the optimum regression equation judgment period. Furthermore, in the method for calculating the appropriate inventory quantity according to the present invention, the arithmetic unit calculates the regression equation for calculating the sales volume judgment period predicted value closest to the value in the optimal regression formula judgment period of the sales quantity data as an optimal regression formula. Includes selecting as. Further, in the method for calculating the appropriate inventory amount according to the present invention, the arithmetic unit calculates a sales volume current forecast value that is a forecast value of the sales volume for the current period from the current factor data by the optimal regression equation; Adding a predetermined safety stock quantity to the quantity current forecast value and calculating an appropriate stock quantity. And the appropriate inventory quantity calculation method of this invention includes the said output device outputting the said appropriate inventory quantity.

  In the appropriate inventory amount calculation method of the present invention, each of the plurality of predetermined target periods may have different start periods and the same end periods.

  In the appropriate inventory amount calculation method of the present invention, the storage device stores a correlation coefficient / rank correspondence table (304) that defines the correspondence between the correlation coefficient and the rank of the degree of correlation. In addition, there are a plurality of factors, and the factor data exists corresponding to each of the factors. Then, in the method for calculating an appropriate inventory amount according to the present invention, the arithmetic device is configured to compare the sales quantity data with each of the factor data for the longest target period having the longest period among the plurality of predetermined target periods. Including calculating the number of relationships. Furthermore, the appropriate inventory amount calculation method of the present invention includes that the arithmetic device ranks each of the factor data by referring to the correlation coefficient / rank correspondence table from the correlation coefficient. Furthermore, in the appropriate inventory amount calculation method of the present invention, when the ranking satisfies a predetermined first condition, the factor data to which the arithmetic device is assigned a rank equal to or higher than a predetermined rank is used as the explanatory variable. And calculating the regression equation which is a multiple regression equation.

  When the ranking satisfies the predetermined second condition, the appropriate inventory amount calculation method according to the present invention uses the factor data to which the arithmetic device is assigned the highest rank as the explanatory variable by a single regression equation. Calculating certain regression equations.

  In the appropriate inventory quantity calculation method of the present invention, the storage device stores a first standard deviation / smoothing constant correspondence table (305) that defines the correspondence between the first standard deviation and the smoothing constant. Then, in the method for calculating an appropriate inventory amount according to the present invention, when the ranking does not satisfy both the first condition and the second condition, the arithmetic unit performs a predetermined first target period that ends in the previous period. Calculating the average value of the sales volume first target period data, which is the sales volume data included. Furthermore, in the method for calculating an appropriate inventory amount according to the present invention, the arithmetic unit divides each of the sales quantity first target period data by the sales quantity data one period before the sales quantity first target period data. And calculating sales volume fluctuation data. Furthermore, the appropriate inventory quantity calculation method of the present invention includes that the arithmetic device calculates the first standard deviation which is a standard deviation of the sales volume increase / decrease degree data. Furthermore, the method for calculating the appropriate inventory quantity according to the present invention includes that the arithmetic unit determines the smoothing constant from the first standard deviation with reference to the first standard deviation / smoothing constant correspondence table. Furthermore, in the method for calculating the appropriate inventory amount according to the present invention, instead of calculating the sales volume current period forecast value by the optimal regression equation, the arithmetic device calculates the determined smoothing constant, the average value, and the previous period. And calculating the current sales volume forecast value from the sales volume data by an index smoothing formula.

  In the method of calculating an appropriate inventory amount according to the present invention, the calculation unit calculates the sales quantity from the factor data included in the second target period, which is the target period for which the optimal regression formula is calculated, by the optimal regression formula. Calculating time-series data of sales volume second target period predicted value, which is a predicted value in the second target period. Furthermore, in the method for calculating an appropriate inventory quantity according to the present invention, the arithmetic device takes a difference between time series data of the sales volume second target period predicted value and time series data of the sales quantity data in the second target period. This includes calculating the time series data of the regression equation prediction error which is the prediction error of the prediction by the regression equation. Furthermore, in the method for calculating an appropriate inventory amount according to the present invention, the arithmetic device calculates a second standard deviation that is a standard deviation of time series data of the regression equation prediction error, and a predetermined constant is set for the second standard deviation. And calculating the safety stock amount.

  In the method for calculating an appropriate inventory amount according to the present invention, the arithmetic unit is configured to calculate the first target period of the sales volume by the exponential smoothing formula from the smoothing constant, the average value, and the sales volume first target period data. Calculating the time series data of the sales volume first target period prediction value, which is the prediction value at. Furthermore, in the method for calculating an appropriate inventory amount according to the present invention, the arithmetic device takes a difference between time series data of the sales volume first target period prediction value and time series data of the sales quantity data in the first target period. This includes calculating time series data of an exponential smoothing type prediction error which is a prediction error of prediction by the exponential smoothing type. Further, according to the method for calculating an appropriate inventory amount of the present invention, the arithmetic device calculates a third standard deviation which is a standard deviation of time series data of the exponential smoothing type prediction error, and the third standard deviation is predetermined. The safety stock quantity is calculated by multiplying by a constant.

  In the appropriate inventory amount calculation method of the present invention, each of the plurality of predetermined target periods has different start times and the same end time. In the method of calculating an appropriate inventory amount according to the present invention, when the ranking satisfies the first condition, the arithmetic device is a standard deviation of the sales quantity data included in the longest target period. Including calculating the standard deviation. Furthermore, in the method for calculating an appropriate inventory amount according to the present invention, the arithmetic device adds the fourth standard deviation to each of the time series data of the sales volume data included in the longest target period, and at the time of the first upper limit value. Including calculating series data. Furthermore, in the method for calculating an appropriate inventory amount according to the present invention, the arithmetic unit subtracts the fourth standard deviation from each of the time series data of the sales volume data included in the longest target period, and at the time of the first lower limit value. Including calculating series data. Furthermore, the appropriate inventory amount calculation method of the present invention is included in a fourth target period, which is the target period in which each of the multiple regression equations is calculated by the arithmetic device using each of the multiple regression equations. Calculating sales volume fourth target period predicted value time series data, which is time series data of predicted values of the sales quantity in the fourth target period from the factor data. Furthermore, the appropriate inventory amount calculation method of the present invention is smaller than the first upper limit value time-series data in all periods in which the sales volume fourth target period predicted value time-series data is included in the fourth target period. And when there is no multiple regression equation that satisfies that and greater than the first lower limit time-series data, and the ranking satisfies the second condition, the computing device comprises: And calculating the regression equation, which is a simple regression equation, using the factor data assigned the highest rank as the explanatory variable.

  In the appropriate inventory amount calculation method of the present invention, each of the plurality of predetermined target periods has different start times and the same end time. In the case where the ranking satisfies the second condition, the appropriate inventory amount calculation method according to the present invention is such that the arithmetic device calculates a fifth standard deviation that is a standard deviation of the sales quantity data in the longest target period. Including doing. Furthermore, the appropriate inventory amount calculation method of the present invention is included in the fifth target period, which is the target period in which the arithmetic device is calculating the optimal single regression equation that is the optimal regression equation as a single regression equation. The second upper limit time series data is calculated by adding the fifth standard deviation to each of the time series data of the sales volume data. Furthermore, in the method for calculating an appropriate inventory amount according to the present invention, the arithmetic unit subtracts the fifth standard deviation from each of the time series data of the sales volume data in the fifth target period to obtain the second lower limit time series data. Including calculating. Furthermore, in the method for calculating an appropriate inventory amount according to the present invention, the arithmetic unit calculates the predicted value of the sales quantity in the fifth target period from the factor data included in the fifth target period by the optimal single regression equation. It includes calculating the sales volume fifth target period predicted value time series data which is time series data. Furthermore, the appropriate inventory amount calculation method of the present invention is smaller than the second upper limit value time-series data in all periods in which the sales volume fifth target period predicted value time-series data is included in the fifth target period. And the second lower limit time-series data are not satisfied, the calculation unit is a sales quantity that is the sales quantity data included in a predetermined first target period that ends in the previous period. Calculating the average value of the first target period data. Furthermore, in the method for calculating an appropriate inventory quantity according to the present invention, the arithmetic unit divides each of the sales quantity first target period data by the sales quantity data one period before the sales quantity first target period data. And calculating the sales volume increase / decrease degree data. Furthermore, in the method for calculating an appropriate inventory amount according to the present invention, the arithmetic device calculates the first standard deviation, which is a standard deviation of the sales volume increase / decrease degree data, and the first standard deviation from the first standard deviation. And determining the smoothing constant with reference to the smoothing constant correspondence table. Furthermore, in the method for calculating an appropriate inventory amount according to the present invention, the arithmetic unit replaces the determined smoothing constant, the average value, and the previous month with the optimal regression equation instead of calculating the sales volume current period forecast value. And calculating the sales volume current period forecast value from the sales volume data by the index smoothing formula.

  According to the present invention, there is provided an appropriate inventory quantity calculation system and an appropriate inventory calculation method for calculating an appropriate inventory quantity of a product based on past time-series data, and calculating an appropriate inventory quantity in consideration of a time change of product characteristics. An appropriate inventory quantity calculation system and an appropriate inventory quantity calculation method are provided.

  In addition, according to the present invention, an appropriate inventory quantity calculation system and an appropriate inventory quantity calculation method for calculating an appropriate inventory quantity by selecting an optimal sales quantity prediction method for each product are provided.

  Furthermore, according to the present invention, there is provided an appropriate inventory quantity calculation system and an appropriate inventory quantity calculation method capable of calculating an appropriate inventory quantity with high prediction accuracy even for a product whose sales volume varies greatly with time.

  With reference to the attached drawings, the best mode for carrying out an appropriate inventory quantity calculation system and an appropriate inventory quantity calculation method according to the present invention will be described below.

(First embodiment)
FIG. 1 shows the configuration of an appropriate inventory quantity calculation system according to an embodiment of the present invention. The appropriate inventory quantity calculation system of the present invention includes a store PC 1, an application server 2, a data center server 3, a POS terminal 4, and a representative PC 5. These are connected by a communication network 6. The store PC 1 is a computer installed in a store of a retail chain. The store PC 1 includes a keyboard 11 and a display device 12. The POS terminal 4 is a computer installed at a cash register of a retail chain store, and includes a barcode reader 41. The representative PC 5 is a computer installed at the headquarters of the retail chain.

  The configuration of the application server 2 is shown in FIG. The application server 2 includes a central processing unit 21, a communication device 22, and a storage device 23. The communication device 22 communicates with the store PC 1 and the data center server 3. The storage device 23 stores an ordering work screen editing program 210. The central processing unit 21 generates an ordering work screen according to the ordering work screen editing program 210.

  The configuration of the data center server 3 is shown in FIG. The data center server 3 includes a central processing unit 31, a communication device 32, and a storage device 33. The communication device 32 communicates with the application server 2, the POS terminal 4, and the representative terminal 5. The storage device 33 includes a database of a product master 301, a sales record master 302, an environment master 303, a correlation coefficient / rank correspondence table 304, a standard deviation / smoothing constant correspondence table 305, and a monthly sales record table 306, and an appropriate inventory. An amount calculation program 310 is stored. The central processing unit 31 executes a sales record update process S1, a sales information registration process S2, an environment information registration process S3, and an appropriate inventory quantity calculation process S4, which will be described later, in accordance with the appropriate inventory quantity calculation program 310.

  The data structure of the product master 301, the sales record master 302, the environment master 303, and the monthly sales record table 306 is shown in FIG.

  The merchandise master 301 uses “store code”, “barcode type”, and “barcode” as main keys, and attributes “first shipment date”, “sales unit price”, “appropriate inventory”, and “number of advertisements”. Have as. Here, “bar code type” indicates the type of bar code such as JAN, and “bar code” indicates a numerical value represented by the bar code. Which product of which store is specified by a combination of “store code”, “bar code type”, and “bar code”.

  The sales record master 302 uses “store code”, “recording date”, “barcode type”, and “barcode” as main keys, and “sales quantity”, “sales unit price”, “weather”, “temperature” ”And“ Advertisement presence / absence ”as attributes. Here, “weather” is a numerical value representing the weather, such as 1 for clear weather, 0 for cloudy, and −1 for rain. Since the record of the sales record master 302 is the original data for predicting the sales volume, it is desirable to have a sales record of the past one year or more. Therefore, it is desirable that the data center server 13 has a large storage capacity. If the storage capacity of the data center server 13 is limited, the sales performance data may be converted into text data and compressed and stored, and decompressed when calculating the appropriate inventory amount.

  The environment master 303 has “store code” as a main key and “weather prediction value” and “temperature prediction value” as attributes. The record of the environment master 303 is one record per store, and is prediction data regarding the sales environment of the current month.

  The monthly sales performance table 306 has “store code”, “recording date”, “barcode type”, and “barcode” as the main keys, and “sales volume”, “unit price”, “elapsed after initial shipment” It has “Year / Month”, “Weather”, “Average temperature”, and “Number of advertisements” as attributes. The monthly sales record table 306 is created based on the product master 301 and the sales record master 302. In the monthly sales performance table 306, “Sales volume” is the total value of the sales volume for the month (recording month), “Sales unit price” is the average value of the sales unit price for that month, “After the first shipment” `` Elapsed date '' is the year and month since the product was shipped for the first time, `` Weather '' is the average value for the month in terms of weather, `` Average temperature '' is the average temperature for that month, “Number of advertisements” is the total number of days on which advertisements were made in the month.

  Next, an outline of processing executed by the appropriate inventory quantity calculation system according to the present invention will be described with reference to FIG. At each store, orders for the current month are made once a month. Prior to the ordering operation, the data center server 3 calculates an appropriate inventory amount using sales performance data up to the previous month.

  The appropriate inventory quantity calculation system according to the present invention can also be applied to the case where the ordering operation is performed during the month, twice a month, or weekly. For example, when ordering is performed on the 16th of every month, when placing an order on May 16, the month from May 16 to June 15 is the current month, the period from April 16 to May 15 is the previous month, March The period from the 16th to the 15th of April may be the month before the other day.

  When the date changes from the last day of the previous month to the first day of the current month, the data center server 3 performs a sales record update process S1, a sales information registration process S2, and an environment information registration process S3.

  In the sales record update process S <b> 1, the sales volume, unit sales price, weather, temperature, and advertisement presence / absence data recorded from the POS terminal 4 are added to the sales record master 302. The sales record update process S1 is executed daily as a night process.

  In the sales information registration process S <b> 2, “Sales unit price” and “Number of advertisements” of the product master 301 are updated based on the schedule of the current month input from the representative PC 5. The sales information registration process S2 needs to be performed prior to the appropriate inventory amount calculation process S4 described later, but may be performed in the previous month.

  In the environment information registration process S3, “weather forecast value” and “temperature forecast value” in the environment master 303 are updated. The “weather forecast value” is updated based on the forecast value of the current month's weather input from the representative PC 5. Here, the predicted value of the weather for this month is determined based on the prediction of a public organization such as the Japan Meteorological Agency or a reliable organization such as a private weather information provider. The “temperature predicted value” is updated based on the predicted value of the average temperature of the current month input from the representative PC 5. Here, the predicted value of the average temperature for the current month is determined based on statistical information for the past several years or based on reliable agency predictions.

  The data center server 3 performs the appropriate inventory amount calculation process S4 after performing the sales record update process S1, the sales information registration process S2, and the environment information registration process S3. The data center server 3 calculates the appropriate inventory amount for the current month based on the product master 301, the sales record master 302, and the environment master 303 in the appropriate inventory amount calculation process S <b> 4 and stores it in the product master 301.

  Next, each process for ordering work will be described. The person in charge of ordering at each store inputs a command for displaying an ordering work screen from the keyboard 11 of the store PC 1 (input processing S10).

  In response to the input process S10, the store PC 1 requests an ordering work screen from the application server 2 (ordering work screen request process S11). At this time, store PC1 transmits the store code of the store where store PC1 is installed. After the ordering work screen request process S11, the store PC 1 is in a waiting state until the ordering work screen response process S15 is completed.

  In response to the ordering work screen request process S11, the application server 2 requests the data center server 3 for an appropriate inventory quantity for each product regarding the store where the shop PC 1 is installed (appropriate inventory quantity request process S12). At this time, the application server 2 transfers the store code.

  In response to the appropriate stock quantity request processing S12, the data center server 3 acquires the appropriate stock quantity of the product handled in the store where the store PC 1 is installed from the product master 301 using the store code as a clue, and the application server 2 (appropriate inventory amount response process S13).

  In response to the appropriate inventory quantity response process S13, the application server 2 generates an ordering work screen based on the appropriate inventory quantity received from the data center server 3 (ordering work screen editing process S14). Here, the ordering operation screen includes information on the appropriate inventory amount for each product.

  Subsequently, the application server 2 sends the generated ordering work screen data to the store PC 1 (ordering work screen response process S15).

  In response to the ordering work screen response process S15, the store PC 1 displays the ordering work screen on the display device 12 (ordering work screen display process S16). The person in charge of ordering determines the order quantity of each product by subtracting the current inventory quantity from the displayed appropriate inventory quantity.

  The appropriate inventory amount calculation process S4 will be described in detail with reference to the flowcharts of FIGS.

  As shown in FIG. 6, the data center server 3 selects a prediction formula for predicting the sales volume for the current month (step S400), and predicts the sales volume based on the selected prediction formula (steps S410, S420, S430). The safety stock quantity is calculated (step S440), and the proper stock quantity calculation process S4 is executed by calculating the proper stock quantity (step S450). Here, in step S400, a prediction formula for predicting the sales quantity is selected, and the sales quantity is predicted by a multiple regression equation (step S410), the sales quantity is predicted by a single regression equation (step S420), or The processing is handed over to the sales volume prediction by the exponential smoothing formula (step S430). In step S410, a multiple regression equation is obtained, and the sales quantity is predicted by the obtained multiple regression equation. If it is not appropriate to predict the sales quantity by the multiple regression equation, the process is passed to step S420. In step S420, a single regression equation is obtained and the sales quantity is predicted by the obtained single regression equation. However, if it is not appropriate to predict the sales quantity by the single regression equation, the process is transferred to step S430. In step S430, an exponential smoothing equation is obtained, and the sales volume is predicted by the obtained exponential smoothing equation. In step S450, the sales amount predicted in step S410, S420, or S430 and the safety stock amount calculated in step S440 are summed to calculate an appropriate inventory amount.

  The data center server 3 executes step S4 for each product at each store to calculate an appropriate inventory amount.

  Details of step S400 for selecting the prediction formula will be described with reference to FIG. Step S400 includes steps S401 to 407.

  First, based on the product master 301 and the sales record master 302, monthly sales for each product of each store, total sales volume, sales unit price, elapsed time after first shipment, weather, average temperature, and number of advertisements are calculated monthly. An achievement table 306 is created and stored in the storage device 33 (step S401). The monthly sales record table 306 may be saved on the memory when the capacity of the system memory of the data center server 3 is sufficiently large.

  Next, based on the monthly sales results table 306, the sales quantity is an objective variable, the elapsed time after the first shipment, the sales unit price, the weather, the average temperature, and the number of advertisements are explanatory variables, and the correlation coefficient for each explanatory variable. Is calculated (step S402). For the calculation of the correlation coefficient, data in a predetermined period Rmax is used. As shown in FIG. 12, if the current month is K month, the start of the period Rmax is (K-2) -Lmax month, and the end is K-2 month (previous month). The length of the period Rmax is predetermined as 5 years or the like.

  Next, A, B, and C are ranked for each explanatory variable with reference to the correlation coefficient / rank correspondence table 304 shown in FIG. 13 (step S403). Here, when the absolute value of the correlation coefficient is 0.7 or more and 1 or less, it is A rank, when it is 0.4 or more and less than 0.7, it is B rank, and 0.0 or more and less than 0.4. In the case of C rank.

  Next, a variance expansion coefficient is calculated for each explanatory variable in the monthly sales performance table 306 (step S404).

  Next, when there are a plurality of explanatory variables whose rank is A or B and whose variance expansion coefficient exceeds a predetermined value (for example, 10), the highest correlation coefficient among them. The rank of the explanatory variable excluding the explanatory variable indicating “” is corrected to the lowest C rank (step S405). Step S405 prevents the multiple regression equation from being correctly calculated when there is a combination having a strong correlation between the explanatory variables.

  Next, when the condition A is satisfied, the process proceeds to step S410, and otherwise, the process proceeds to step S407 (step S406). Here, the condition A means that there are two or more explanatory variables of rank B or higher. When the process proceeds to step S407, if the condition B is satisfied, the process proceeds to step S420. Otherwise, the process proceeds to step S430. Here, the condition B means that there is only one explanatory variable of rank A, or that there is only one explanatory variable of rank B without any explanatory variable of rank A.

  The appropriate inventory amount calculation system according to the present invention selects an optimal prediction formula for each product in step S400. Therefore, since the prediction considering the product characteristics is made, it is possible to predict the sales quantity with high accuracy.

  Details of step S410 for predicting the sales volume by the multiple regression equation will be described with reference to FIGS. Step S410 includes steps S411 to 416.

  First, in the monthly sales result table 306, a target period R1 for calculating a multiple regression equation is selected (step S411). As shown in FIG. 12, the start period of the target period R1 is (K-2) -LJanuary, and the final period is K-2 month (previous month). L1 is an integer from Lmin to Lmax. The length of the target period R1 can take various lengths in a range between a predetermined shortest length and a longest length (for example, 6 months and 5 years). Here, when the target period R1 has the longest length, the target period R1 matches the period Rmax. When the target period R1 has the shortest length, the start period of R1 is (K-2) -Lmin months.

  Next, a multiple regression equation is obtained for the selected target period (step S412). In calculating the multiple regression equation, the constant term and the partial regression coefficient are determined by the least square method. As an explanatory variable for obtaining a multiple regression equation, an explanatory variable whose rank is set to A or B after the rank correction in step S405 is used. The actual sales volume is used as the objective variable.

  In the next step S413, multiple regression equations are obtained for all cases from when the start of the target period R1 is (K-2) -Lmin month to (K-2) -Lmax month. Sometimes the process proceeds to step S414, and if not, step S411 and step S412 are repeated to obtain a multiple regression equation.

Next, the validity of each of the multiple regression equations obtained by changing the target period R1 is determined (step S414). The validity is determined according to the following procedure. First, the predicted value ρN of the sales quantity in N months included in the target period R1 for which this multiple regression equation is obtained is calculated using the multiple regression equation that is the object of determination. Next, the standard deviation δ1 of the actual sales quantity is calculated for the longest period among the target periods R1 used to obtain the multiple regression equation. And it is investigated whether Formula (1) is materialized in all the months included in object period R1.
ωN−δ1 <ρN <ωN + δ1 (1)

  Here, ωN is the actual sales volume in N month. If equation (1) holds for all months within the target period, it is determined that the multiple regression equation is valid, and otherwise, it is determined that the multiple regression equation is not valid. If there is at least one multiple regression equation determined to be valid, the process proceeds to step S415, and if not, the process proceeds to step S420.

  In step S415, an optimal multiple regression equation is selected. When there is one multiple regression equation determined to be valid in step S414, the multiple regression equation is set as the optimum multiple regression equation. When there are a plurality of multiple regression equations determined to be appropriate, the forecast value of the sales volume in K-1 month (previous month) is calculated using the explanatory variable data in the previous month, and is closest to the actual sales volume in the previous month. The multiple regression equation that gives the predicted value is selected as the optimal multiple regression equation.

  In the next step S416, the predicted value of the sales quantity in K month (current month) is calculated by the optimum multiple regression equation selected in step S415. At this time, the data of the product master 301 and the environment master 303 are used as explanatory variables for the current month.

  Next, details of step S420 for predicting the sales quantity by a single regression equation will be described with reference to FIG. Step S420 includes Steps S421 to Step 427.

  In step S421, if the condition B is satisfied, the process proceeds to step S422, and if not, the process proceeds to step S430. If it is determined in step S414 that there is no valid multiple regression equation and the process is delivered, data for which it is not appropriate to use the single regression equation may be delivered. This is because the sales quantity is predicted by the chemical formula (step S430).

  Next, the target period R1 of data used for calculation of the single regression equation is selected (step S422). As shown in FIG. 12, the start period of the target period R1 is (K-2) -LJanuary, and the final period is K-March (previous month). L1 is an integer from Lmin to Lmax. The length of the target period R1 can take various lengths in a range between a predetermined shortest length and a longest length (for example, 6 months and 5 years). Here, when the target period R1 has the longest length, the target period R1 matches the period Rmax. When the target period R1 has the shortest length, the start period of R1 is (K-2) -Lmin months.

  Next, a single regression equation is obtained for the selected target period (step S423). In calculating the simple regression equation, the constant term and the regression coefficient are determined by the least square method. As the explanatory variable for obtaining the simple regression equation, the explanatory variable assigned the highest rank after the rank correction in step S405 is used. The actual sales volume is used as the objective variable.

  In the next step S424, single regression equations are obtained for all cases from when the start of the target period R1 is (K-2) -Lmin month to (K-2) -Lmax month. Sometimes the process proceeds to step S425, and if not, step S422 and step S423 are repeated to obtain a single regression equation.

  In step S425, the predicted value of the sales volume in the previous month is calculated by the single regression equation obtained in step S423 using the explanatory variable data in the previous month, and the single regression equation that makes the prediction closest to the actual sales volume in the previous month is optimal. As a simple regression equation. At this time, the correlation coefficient between the explanatory variable and the objective variable is calculated for each target period, and ranking is performed based on the correlation coefficient / rank correspondence table 304. If the rank is lower than the rank assigned in step S402, the single regression equation obtained for the data for the target period is excluded from the target for obtaining the optimum single regression equation. This is to exclude a simple regression equation in which only the forecast of the sales volume of the previous month coincides with the sales volume performance.

  In step S426, the validity of the optimal single regression equation selected in step S425 is determined. The determination of validity is the same as in step S414. If the formula (1) holds for all the months in the target period, it is determined that the single regression equation is valid, and the process proceeds to step S427. Otherwise, the single regression equation is not valid. And the process proceeds to step S430.

  In the next step S427, the predicted value of the sales volume for the current month is calculated by the optimal single regression equation selected in step S425. At this time, the data in the product master 301 and the environment master 303 is used as the explanatory variable data for the current month.

  In the present invention, as seen in steps S411 to S413 and steps S422 to S424, the target period is changed by moving the start period back and forth, and the optimal regression equation obtained from the data for each target period is selected. The sales volume is predicted by selecting the item. Therefore, in the present invention, it is possible to predict the sales quantity with high accuracy even for a product whose product characteristics change with time.

  Next, details of step S430 for predicting the sales volume by the exponential smoothing formula will be described with reference to FIGS. Step S430 includes steps S431 to 435.

In steps S431 to S434, an exponential smoothing equation is obtained. The exponential smoothing equation is expressed by equation (2).
ρ1 = (1−α) μ + αω (2)

  Here, ρ1 is a predicted value of sales volume in K month (current month), μ is an average value of sales volume results within a predetermined period R2, and ω is a sales volume result in K-1 month (previous month). Here, the period R2 is a period starting from (K-1) -LFebruary and ending in K-1 month (previous month). The smoothing constant α is a numerical value from 0 to 1 obtained in step S434.

  In general, in the exponential smoothing formula, if ρ1 is the predicted value of the current month and ω is the actual value of the previous month, μ is the predicted value of the previous month. However, in the present invention, the exponential smoothing formula is as described above. Shall be included.

  First, in the monthly sales result table 306, the sales quantity results of each month included in the period R2 are divided by the average value of the sales quantity results in the period R2, and the sales quantity results are standardized (step S431). Subsequently, the standardized sales volume result is divided by the standardized sales volume result of the previous month to determine the degree of increase / decrease, and the absolute value of the degree of increase / decrease is further obtained (step S432). Subsequently, the standard deviation of the absolute value of the degree of increase / decrease is calculated (step S433). Subsequently, the smoothing constant α is obtained from the standard deviation with reference to the standard deviation / smoothing constant correspondence table 305 shown in FIG. 14 (step S434). Since standardization is performed, the smoothing constant α can be obtained without being affected by the scale of the sales volume of the product.

  In step S435, the average value in the period R2 of the sales volume results obtained by referring to the monthly sales results table 306 and the sales volume results in the previous month are substituted into the formula (2), and the predicted value of the sales volume in the current month. Is calculated.

  In step S440, a safety stock quantity is calculated. The safety stock quantity is calculated according to the procedures of steps S441 and S442 shown in FIG.

  In step S441, with reference to the monthly sales performance table 306, a prediction error is calculated for the period used to predict the sales quantity, and a standard deviation δ2 of the prediction error is obtained. In the case where the prediction is made by the multiple regression formula or the single regression formula, the optimum multiple regression formula or the optimum single regression formula and the optimum multiple regression formula or the optimum based on the explanatory variable data of the monthly sales performance table 306 are calculated. A forecast value of sales volume is calculated for the target period for which a simple regression equation is being calculated, and a prediction error is obtained by taking the difference between the forecast value and the actual sales volume. In the case where the prediction is performed using the exponential smoothing formula, the forecast value of the sales quantity is calculated for the period R2 using the sales volume actual data in the monthly sales performance table 306 and the exponential smoothing formula. The prediction error is obtained by taking the difference from the actual sales volume. For example, by using the sales volume record in the month X as ω in the equation (2), the sales volume record in the next month of the month X can be obtained as ρ1.

Subsequently, in step S442, the safety stock quantity φ is obtained by Expression (3).
φ = βδ2 (3)

  Here, β is a safety factor. Generally, 1.65 or 2.33 is used as the safety factor β. When the safety factor β is 1.65, the risk factor is 5%, and when it is 2.33, the risk factor is 1%.

  When obtaining the safety stock amount, the lead time (the period from when the product is ordered until it is delivered) may be taken into consideration. In this case, the safety stock quantity is obtained by adding the predicted value of the sales quantity during the lead time period. For example, in the case where the predicted value of the sales volume of a certain product for the current month is 100 pieces and the lead time is 0.5 month, 0.5 × 100 pieces = 50 pieces are added, so the safety stock quantity φ is βδ2 + 50 pieces. It becomes.

  In the present invention, as can be seen in step S440, the safety stock amount is calculated based on the standard deviation of the prediction error in the past certain period. In other words, in the present invention, since the safety stock quantity is calculated according to the reliability of the prediction, it is possible to prevent the excess inventory from being held. On the other hand, when calculating the safety stock based on the standard deviation of the actual sales volume over a certain period in the past, that is, when calculating the safety stock according to the severity of the sales volume fluctuation, For products that have large fluctuations but can predict the sales volume, a large amount of safety stock is calculated, so it is easy to have excess inventory. For example, the superiority of the present invention is clear if a disposable body warmer or the like that can predict the sales volume from the temperature and weather, although the sales volume varies greatly depending on the season, is clear.

In step S450, the appropriate inventory quantity σ is obtained from the equation (4), and the obtained appropriate inventory quantity σ is stored in the product master 301.
σ = ρ2 + φ (4)

  Here, ρ2 is the predicted value of the sales volume for the current month obtained in step S410, step S420, or step S430.

(Explanation using specific data on the procedure for calculating the appropriate amount of inventory)
Next, the procedure for calculating the appropriate inventory quantity executed in the appropriate inventory quantity calculation system of the present invention will be described using specific data.

  A specific example 306 'of the monthly sales record table 306 is shown in FIG. In FIG. 15, the products A, B, and C are products specified by the store code, the barcode type, and the barcode of the sales record master 302. FIG. 15 shows the sales performance (sales volume), sales unit price, average temperature, and elapsed time (elapsed month after first shipment) for each of the products A to C. Here, the sales performance is the total value of the sales performance in each recording date, the sales unit price and the average temperature are the average values in each recording date, and these are calculated from the data of the sales performance master 302 in step S401. Is. The elapsed date is calculated in step S401 from the initial shipment date of each product shown in the specific example 301 'of the product master 301 shown in FIG. 16 and the date of recording in FIG.

  For each of the products A, B, and C, FIG. 17 shows the elapsed time, the sales unit price, the correlation coefficient for the average temperature, the variance expansion coefficient, and the rank. Here, the correlation coefficient is calculated by comparing the sales record in the longest target period Rmax in which the recording date in FIG. 15 is from December 2004 to November 2005, the elapsed time, the sales unit price, and the average temperature. Correlation coefficient, calculated in step S402. The dispersion expansion coefficient is calculated in step S404 from the elapsed time, the sales unit price, and the average temperature in the period from December 2004 to November 2005. The ranks are ranked in step S403 based on the correlation coefficient / rank correspondence table 304. In FIG. 17, since the dispersion expansion coefficient does not exceed a predetermined value (= 10), the rank is not corrected (step S405).

  For the product A, the elapsed time, the sales unit price, and the average temperature are all ranked A, so the condition A is satisfied, and the process is handed over to step S410 (step S406). For the product B, the elapsed time and the average temperature are ranked C, and the sales unit price is ranked A, so the condition A is not satisfied and the condition B is satisfied. The process is handed over to (Steps S406 and S407). For the product C, the elapsed year, the sales unit price, and the average temperature are all ranked C, so both the conditions A and B are not satisfied, and the process is handed over to step S430 (steps S406 and S407).

  FIG. 18 shows a table of sales results, the lower limit value, the upper limit value, the sales forecast, and the prediction error for the product A in each recording date. Here, the sales period is an objective variable, the sales unit price, average temperature, and elapsed time are explanatory variables, and the target period is changed from the longest target period (December 2004 to November 2005) to the shortest target period (June 2005). To November 2005) to obtain a multiple regression equation (steps S411 to S413). In FIG. 18, the target period is selected every two months, but may be selected every month, and the number of months for which the target period is selected can be arbitrarily set. Sales forecasts calculated by the multiple regression equation obtained for the target period from December 2004 to November 2005 are shown in columns 2004/12 to 2005/11. For example, the sales forecast for December 2004 shown in FIG. 18 is 97 by substituting the sales unit price, the average temperature, and the elapsed time of the product A for December 2004 in the multiple regression equation as shown in FIG. Calculated as 93. Sales forecasts for other target periods are calculated in the same manner. Since the standard deviation δ1 in the longest target period Rmax (December 2004 to November 2005) in FIG. 18 is 14.5, the lower limit value (ωN−δ1) and the upper limit value (ωN + δ1) are, for example, 2004. For December, the lower limit is calculated as 87.5 (= 102-14.5), and the upper limit is calculated as 116.5 (= 102 + 14.5). In FIG. 18, since all the sales forecasts in the four target periods are between the upper limit value and the lower limit value in the recording date, it is determined that there is an appropriate multiple regression equation in step S414, and the process proceeds to step S415. Is delivered. Looking at the sales forecast in the previous month (December 2005), the sales forecast (124.19) when the target period is from February 2005 to November 2005 is the closest to the sales record (124). Therefore, the multiple regression equation obtained by setting the target period from February 2005 to November 2005 is selected as the optimum multiple regression equation (step S415). And the sales forecast for this month is 118.10 by inputting the unit price of 97 yen for this month (January 2006), the predicted temperature of 4.5 ° C, and the past 14 months into the optimal multiple regression equation. Calculated (step S416).

  In the target period (February 2005-November 2005) for which the optimum multiple regression equation is obtained, the forecast error is calculated by subtracting the actual sales from the sales forecast, and further, the standard deviation δ2 of the forecast error is calculated. .58 (step S441). When 1.65 is used as the safety factor β, the safety stock quantity φ is calculated to be 5.91 (step S442). Therefore, the appropriate inventory quantity is calculated as 125 pieces by rounding up the decimal point of 118.10 + 5.91 = 124.01 (step S450).

  FIG. 19 shows a table of sales results, the lower limit value, the upper limit value, the sales forecast, and the prediction error for the product B in each recording date. Here, the sales period is the objective variable, the sales unit price is the explanatory variable, and the target period changes from the longest target period (December 2004 to November 2005) to the shortest target period (June 2005 to November 2005). Thus, a single regression equation is obtained (steps S422 to S424). In FIG. 19, the target period is selected every two months, but may be selected every month, and the number of months for which the target period is selected can be arbitrarily set. Here, for the four target periods shown in FIG. 19, the correlation coefficient between the sales unit price and the sales record shown in FIG. 15 and the ranks ranked with reference to the correlation coefficient / rank correspondence table 304 are shown in FIG. Shown in As shown in FIG. 20, since the rank is A rank in all the four target periods, there is no case where it is lower than the rank given in step S402 (rank for the sales unit price for the product B shown in FIG. 17). . Therefore, the single regression equation obtained for all the target periods is the target for selecting the optimum single regression equation. In FIG. 19, sales predictions calculated by a single regression equation obtained for the target period from December 2004 to November 2005 are shown in columns 2004/12 to 2005/11. For example, by substituting the unit sales price of 98 yen for product B in December 2004 as shown in FIG. 15 into a simple regression equation, the sales forecast in December 2004 as shown in FIG. 19 is 111.33. Is calculated. Sales forecasts for other target periods are calculated in the same manner. Since the standard deviation δ1 in the longest target period Rmax (December 2004 to November 2005) in FIG. 19 is 14.8, the lower limit (ωN−δ1) and the upper limit (ωN + δ1) are, for example, 2004. For December, the lower limit is calculated to be 85.2 (= 100-14.8), and the upper limit is calculated to be 114.8 (= 100 + 14.8). Looking at the sales forecast for the previous month (December 2005), the sales forecast (108.17) when the target period is from December 2004 to November 2005 is the closest to the sales record (103). Therefore, the single regression equation obtained by setting the target period from December 2004 to November 2005 is selected as the optimal single regression equation (step S425). In FIG. 19, since the sales forecast by the optimal single regression equation is between the upper limit value and the lower limit value in all the months of the target period for which the optimal single regression equation is obtained, it is determined that it is appropriate in step S426. The process is handed over to step S427. Then, the sales forecast for the current month (January 2006) is calculated to be 120.83 by inputting the unit price of 92 yen (see FIG. 16) in the optimal single regression equation (step S427).

  In the target period (December 2004 to November 2005) for which the optimal single regression equation is obtained, the prediction error is calculated by subtracting the sales result from the sales forecast, and further, the standard deviation δ2 of the prediction error is calculated as 8 .32 (step S441). If 1.65 is used as the safety factor β, the safety stock quantity φ is calculated to be 13.73 (step S442). Therefore, the appropriate inventory amount is calculated as 135 pieces by rounding up the decimal point of 120.83 + 13.73 = 134.56 (step S450).

  FIG. 21 shows the sales results, the standardized sales results, the degree of increase / decrease, the sales prediction, and the prediction error for the product C in each recording date. In this specific example, the period from January 2005 to December 2005 corresponds to the target period R2. The standardized sales performance is calculated by dividing the sales performance of each recording date by 105.58, which is the average value μ of the sales performance in the target period R2 (step S431). Then, the degree of increase / decrease is calculated by dividing the standardized sales performance of the month by the standardized sales performance of the previous month (step S432). The degree of increase / decrease can also be calculated by dividing the sales record for the month by the sales record for the previous month. In this case, step S431 and step S432 can be performed in one step. Since the standard deviation of the increase / decrease data in the target period R2 is 0.041, the smoothing constant α is determined to be 0.6 from the standard deviation / smoothing constant correspondence table 305 (step S434). Here, if the smoothing constant α is 0.6, the average value μ of the sales results in the target period R2 is 105.58, and the sales volume result ω in the previous month (December 2005) is 103, the formula (2 ) To calculate 104.03 sales forecasts for the current month (January 2006) (step S435).

  In the target period R2 (January 2005-December 2005), the forecast error is calculated by subtracting the actual sales from the sales forecast, and the standard deviation δ2 of the forecast error is 3.51 (step) S441). When 1.65 is used as the safety factor β, the safety stock quantity φ is calculated to be 5.79 (step S442). Therefore, the appropriate inventory quantity is calculated to be 110 by rounding up the decimal part of 104.03 + 5.79 = 109.82 (step S450). Note that the sales forecast for the month belonging to the target period R2 is calculated by using the sales performance of the previous month in Equation (2) instead of ω.

(Other embodiments)
In the present invention, the reliability of the appropriate inventory quantity calculation may be displayed together with the appropriate inventory quantity in step S16. For this purpose, the data center server 3 stores the standard deviation δ2 of the prediction error in the product master 301 in step S441. In step S13, the application server 2 acquires the standard deviation δ2. Further, in step S14, the application server 2 generates a standard deviation / reliability correspondence table (not shown) that defines a correspondence relationship between the standard deviation δ2 and the reliability stored in the storage device 23 in advance when generating the ordering work screen. , The reliability such as “high reliability of calculation of appropriate inventory quantity” or “low reliability of calculation of appropriate inventory quantity” corresponding to the standard deviation δ2 is displayed on the ordering work screen. The person in charge of ordering can increase or decrease the order quantity with reference to the reliability of the appropriate inventory quantity calculation displayed on the display device 12.

  In the present invention, by preparing the correlation coefficient / rank correspondence table 304 and the standard deviation / smoothing constant correspondence table 305 for each product, it is possible to calculate the appropriate stock amount according to the characteristics of each product. Further, the correlation coefficient / rank correspondence table 304 and the standard deviation / smoothing constant correspondence table 305 may be a growth type system that corrects the correlation based on past prediction results. For example, in the standard deviation / smoothing constant correspondence table 305 shown in FIG. 14, when the standard deviation is 0.04, the smoothing constant is 0.6. In this method, the constant is corrected from 0.6 to 0.5.

  In the validity judgment of the present invention (steps S414 and S426), the multiple regression equation or the single regression equation satisfying the equation (1) is valid for all the months in the target period R1, but all the target period R1 It may be appropriate when the expression (1) is satisfied not only in the month but also in the previous month.

FIG. 1 is a diagram showing a configuration of an appropriate inventory quantity calculation system according to the present invention. FIG. 2 is a diagram illustrating a configuration of the application server. FIG. 3 is a diagram illustrating the configuration of the data center server. FIG. 4 is a diagram illustrating a data configuration of a product master, a sales record master, an environment master, and a monthly sales record table. FIG. 5 is a time chart showing an outline of processing executed by the appropriate inventory quantity calculation system according to the present invention. FIG. 6 is a flowchart showing the appropriate inventory amount calculation process. FIG. 7 is a flowchart showing prediction formula selection. FIG. 8 is a flowchart showing the prediction of the sales volume by the multiple regression equation. FIG. 9 is a flowchart showing the prediction of sales volume by a single regression equation. FIG. 10 is a flowchart showing the prediction of the sales volume by the exponential smoothing formula. FIG. 11 is a flowchart illustrating the safety stock amount calculation. FIG. 12 is a diagram for explaining a target period for data analysis. FIG. 13 is a diagram showing a correlation coefficient / rank correspondence table. FIG. 14 is a diagram showing a standard deviation / smoothing constant correspondence table. FIG. 15 is a diagram showing a specific example of a monthly sales performance table. FIG. 16 is a diagram illustrating a specific example of the product master. FIG. 17 is a diagram for explaining processing in prediction formula selection. FIG. 18 is a diagram for explaining processing in forecasting sales volume by a multiple regression equation. FIG. 19 is a diagram for explaining processing in the prediction of sales volume by a single regression equation. FIG. 20 is a diagram illustrating changes in ranks of explanatory variables when the target period is changed in the prediction of sales volume by a single regression equation. FIG. 21 is a diagram showing processing in the forecast of sales volume by the exponential smoothing formula.

Explanation of symbols

1 ... Store PC
2 ... Application server 3 ... Data center server 4 ... POS terminal 5 ... Representative PC
6 ... Communication network 11 ... Keyboard 12 ... Display device 21, 31 ... Central processing unit 22, 32 ... Communication device 23, 33 ... Storage device 41 ... Bar code reader 210 ... Ordering work screen editing program 301, 301 '... Product master 302 ... Sales record master 303 ... Environment master 304 ... Correlation coefficient / rank correspondence table 305 ... Standard deviation / smoothing constant correspondence table 306, 306 '... Monthly sales record table 310 ... Appropriate inventory calculation program

Claims (16)

An arithmetic unit;
A storage device;
An input device;
An output device,
The storage device
A time-series database that stores sales volume data that is time-series data of sales volume before the previous period, and factor data that is time-series numerical data of the factors affecting the sales volume before and after the previous period;
A sales performance database that stores both sales volume data source data and factor data source data, both of which are time series data,
A current-period factor database that stores current-term factor data that is numerical data of predicted or planned values of the factors for the current period;
The arithmetic unit is
Storing the sales quantity data source data and the factor data source data input from the input device in the sales record database;
Read the sales volume data source data and the factor data source data from the sales performance database,
The sales volume data source data and the factor data source data are aggregated every period to create the sales volume data and the factor data, respectively.
The created sales quantity data and the factor data are stored in the time series database,
Storing the current period factor data input from the input device in the current period factor database;
Read the sales volume data and the factor data from the time series database,
For each of a plurality of predetermined target periods, a regression equation with the sales quantity data as an objective variable and the factor data as an explanatory variable is calculated,
Read the current period factor data from the current period factor database,
Calculate a sales quantity judgment period forecast value that is a forecast value of the sales quantity in the optimum regression formula judgment period by each of the regression formulas from the factor data in the optimum regression formula judgment period that is a predetermined period before the previous period,
Selecting the regression formula for calculating the sales volume judgment period predicted value closest to the value in the optimal regression formula judgment period of the sales quantity data as an optimal regression formula;
Calculate the sales volume current forecast value, which is the forecast value of the sales volume in the current period, from the current factor data by the optimal regression equation,
Calculate the appropriate inventory quantity by adding a predetermined safety inventory quantity to the sales volume current period forecast value,
The output device outputs an appropriate stock quantity. An appropriate stock quantity calculation system. An appropriate inventory amount calculation system according to claim 1,
Each of the plurality of predetermined target periods has an initial period different from each other and an identical final period. An appropriate inventory amount calculation system according to claim 1,
The storage device stores a correlation coefficient / rank correspondence table that defines the correspondence between the correlation coefficient and the rank of the degree of correlation.
There are multiple factors,
The factor data exists corresponding to each of the plurality of factors,
The arithmetic unit is
Calculating a correlation coefficient between the sales quantity data and each of the factor data for the longest target period of the plurality of predetermined target periods,
Rank each of the factor data with reference to the correlation coefficient / rank correspondence table from the correlation coefficient,
If the ranking satisfies a predetermined first condition,
An appropriate inventory amount calculation system that calculates the regression equation, which is a multiple regression equation, using the factor data with a rank higher than a predetermined rank as the explanatory variable. An appropriate inventory amount calculation system according to claim 3,
If the ranking satisfies a predetermined second condition,
The arithmetic unit is
An appropriate inventory amount calculation system that calculates the regression equation that is a single regression equation using the factor data assigned the highest rank as the explanatory variable. An appropriate inventory amount calculation system according to claim 4,
The storage device stores a first standard deviation / smoothing constant correspondence table that defines a correspondence relationship between the first standard deviation and the smoothing constant;
If the ranking does not satisfy both the first condition and the second condition,
The arithmetic unit is
Calculating the average value of the sales volume first target period data, which is the sales volume data included in the predetermined first target period that ends in the previous period,
Dividing each of the sales volume first target period data by the sales volume data one period before the sales volume first target period data to calculate sales volume change degree data,
Calculating the first standard deviation which is a standard deviation of the sales volume increase / decrease degree data;
Determining the smoothing constant from the first standard deviation with reference to the first standard deviation / smoothing constant correspondence table;
Instead of calculating the sales volume current forecast value by the optimal regression formula, the sales volume current forecast value is calculated from the determined smoothing constant, the average value, and the sales volume data in the previous period by an exponential smoothing formula. Calculate the appropriate inventory quantity calculation system. An appropriate inventory amount calculation system according to claim 1,
The arithmetic unit is
Sales quantity second that is a predicted value of the sales quantity in the second target period by the optimal regression formula from the factor data included in the second target period that is the target period for which the optimal regression formula is calculated. Calculate time series data of the target period forecast value,
Regression prediction error, which is a prediction error of prediction by the optimal regression equation, by taking the difference between the time series data of the sales volume second target period prediction value and the time series data of the sales volume data in the second target period Time series data for
Calculating a second standard deviation which is a standard deviation of time series data of the regression equation prediction error;
An appropriate inventory quantity calculation system that calculates the safety inventory quantity by multiplying the second standard deviation by a predetermined constant. An appropriate inventory amount calculation system according to claim 5,
The arithmetic unit is
Based on the smoothing constant, the average value, and the sales quantity first target period data, the sales quantity first target period predicted value that is the predicted value of the sales quantity in the first target period by the exponential smoothing formula. Calculate time series data,
Exponential smoothing which is a prediction error of prediction by the exponential smoothing formula by taking the difference between the time series data of the sales volume first target period forecast value and the time series data of the sales quantity data in the first target period Calculate time series data of formula prediction error,
Calculating a third standard deviation which is a standard deviation of time series data of the exponential smoothing type prediction error;
An appropriate inventory quantity calculation system that calculates the safety inventory quantity by multiplying the third standard deviation by a predetermined constant. An arithmetic unit;
A storage device;
An input device;
An output device,
The storage device
The time series database stores both the sales volume data that is the previous time series data of the sales volume and the factor data that is the time series numeric data of the factor that affects the sales volume before the previous period. Stores the sales performance database that stores the sales volume data source data and the factor data source data, and the current factor database that stores the current factor data that is the numerical data of the forecasted or planned value of the factor for the current period. An appropriate inventory quantity calculation method executed by an inventory quantity calculation system,
The arithmetic unit is
Storing the sales quantity data source data and the factor data source data input from the input device in the sales record database;
Reading the sales volume data source data and the factor data source data from the sales performance database;
Totaling the sales quantity data source data and the factor data source data for each period to create the sales quantity data and the factor data, respectively;
Storing the generated sales quantity data and the factor data in the time series database;
Storing the current period factor data input from the input device in the current period factor database;
Reading the sales volume data and the factor data from the time series database;
For each of a plurality of predetermined target periods, calculating a regression equation with the sales quantity data as an objective variable and the factor data as an explanatory variable;
Reading the current period factor data from the current period factor database;
Calculating a sales volume judgment period forecast value, which is a forecast value of the sales volume in the optimum regression formula judgment period, from each of the regression formulas from the factor data in the optimum regression formula judgment period which is a predetermined period before the previous period; ,
Selecting the regression equation that calculates the sales volume judgment period prediction value closest to the value in the optimal regression equation judgment period of the sales quantity data as an optimal regression expression;
Calculating a sales volume current forecast value that is a forecast value of the sales volume in the current period from the current factor data by the optimal regression equation;
Adding a predetermined safety stock amount to the sales volume current period forecast value to calculate an appropriate stock amount;
A method for calculating an appropriate inventory quantity, comprising: outputting the appropriate inventory quantity by the output device. A method for calculating an appropriate inventory amount according to claim 8,
Each of the plurality of predetermined target periods has a start period different from each other, and an end period is the same. A method for calculating an appropriate inventory amount according to claim 8,
The storage device stores a correlation coefficient / rank correspondence table that defines the correspondence between the correlation coefficient and the rank of the degree of correlation.
There are multiple factors,
The factor data exists corresponding to each of the plurality of factors,
The arithmetic unit is
Calculating a correlation coefficient between the sales volume data and each of the factor data for the longest target period of the plurality of predetermined target periods,
Ranking each of the factor data with reference to the correlation coefficient / rank correspondence table from the correlation coefficient;
Calculating the regression equation that is a multiple regression equation using the factor data having a rank higher than a predetermined rank as the explanatory variable when the ranking satisfies a predetermined first condition. Inventory calculation method. The method of calculating an appropriate inventory amount according to claim 10,
If the ranking satisfies a predetermined second condition,
A method for calculating an appropriate inventory amount, comprising: calculating the regression equation, which is a single regression equation, using the factor data assigned the highest rank as the explanatory variable. The method for calculating an appropriate inventory amount according to claim 11,
The storage device stores a first standard deviation / smoothing constant correspondence table that defines a correspondence relationship between the first standard deviation and the smoothing constant;
If the ranking does not satisfy both the first condition and the second condition,
The arithmetic unit is
Calculating an average value of sales volume first target period data that is the sales volume data included in a predetermined first target period that ends in the previous period;
Dividing the sales quantity first target period data by the sales quantity data one period before the respective sales quantity first target period data to calculate sales quantity increase / decrease degree data;
Calculating the first standard deviation which is a standard deviation of the sales volume increase / decrease degree data;
Determining the smoothing constant from the first standard deviation with reference to the first standard deviation / smoothing constant correspondence table;
Instead of calculating the sales volume current forecast value by the optimal regression formula, the sales volume current forecast value is calculated from the determined smoothing constant, the average value, and the sales volume data in the previous period by an exponential smoothing formula. A method for calculating an appropriate inventory amount. A method for calculating an appropriate inventory amount according to claim 8,
The arithmetic unit is
Sales quantity second that is a predicted value of the sales quantity in the second target period by the optimal regression formula from the factor data included in the second target period that is the target period for which the optimal regression formula is calculated. Calculating time series data for the target period forecast value,
Regression prediction error, which is a prediction error of prediction by the optimal regression equation, by taking the difference between the time series data of the sales volume second target period prediction value and the time series data of the sales volume data in the second target period Calculating time series data for
Calculating a second standard deviation which is a standard deviation of time series data of the regression equation prediction error;
A method for calculating an appropriate inventory quantity, comprising: calculating the safety inventory quantity by multiplying the second standard deviation by a predetermined constant. A method for calculating an appropriate inventory amount according to claim 12,
The arithmetic unit is
Based on the smoothing constant, the average value, and the sales quantity first target period data, the sales quantity first target period predicted value that is the predicted value of the sales quantity in the first target period by the exponential smoothing formula. Calculating time series data;
Exponential smoothing which is a prediction error of prediction by the exponential smoothing equation by taking the difference between the time series data of the sales volume first target period forecast value and the time series data of the sales quantity data in the first target period Calculating time series data of formula prediction error;
Calculating a third standard deviation which is a standard deviation of time series data of the exponential smoothing type prediction error;
A method for calculating an appropriate inventory quantity, comprising: calculating the safety inventory quantity by multiplying the third standard deviation by a predetermined constant. The method for calculating an appropriate inventory amount according to claim 11,
Each of the plurality of predetermined target periods has different start times and the same end times,
In the case where the ranking satisfies the first condition,
The arithmetic unit is
Calculating a fourth standard deviation which is a standard deviation of the sales volume data included in the longest target period;
Calculating the first upper limit time series data by adding the fourth standard deviation to each of the time series data of the sales volume data included in the longest target period;
Subtracting the fourth standard deviation from each of the time series data of the sales volume data included in the longest target period to calculate first lower limit time series data;
The predicted value of the sales quantity in the fourth target period from the factor data included in the fourth target period, which is the target period for which each of the multiple regression formulas is calculated, by each of the multiple regression formulas Calculating the sales volume fourth target period predicted value time series data, which is time series data of
The sales volume fourth target period predicted value time series data is smaller than the first upper limit time series data and larger than the first lower limit time series data in all periods included in the fourth target period. If there is no multiple regression equation satisfying that, and the ranking satisfies the second condition,
A method for calculating an appropriate inventory amount, comprising: calculating the regression equation, which is a single regression equation, using the factor data assigned the highest rank as the explanatory variable. A method for calculating an appropriate inventory amount according to claim 12,
Each of the plurality of predetermined target periods has different start times and the same end times,
In the case where the ranking satisfies the second condition,
The arithmetic unit is
Calculating a fifth standard deviation that is a standard deviation of the sales volume data in the longest target period;
The fifth standard deviation is included in each of the time series data of the sales volume data included in the fifth target period that is the target period in which the optimal single regression formula that is the optimal regression formula as the single regression formula is calculated. To calculate the second upper limit time series data,
Subtracting the fifth standard deviation from each of the time series data of the sales volume data in the fifth target period to calculate second lower limit time series data;
Sales quantity fifth target period predicted value time series data that is time series data of the predicted value of the sales quantity in the fifth target period from the factor data included in the fifth target period by the optimal single regression equation. To calculate,
The sales volume fifth target period predicted value time series data is smaller than the second upper limit value time series data and larger than the second lower limit value time series data in all periods included in the fifth target period. If you are not satisfied with
Calculating the average value;
Calculating the sales volume variation data;
Calculating the first standard deviation;
Determining the smoothing constant;
Instead of calculating the sales volume current forecast value by the optimal regression equation, the sales volume current forecast value by the exponential smoothing formula from the determined smoothing constant, the average value, and the sales volume data in the previous period. Calculating an appropriate inventory quantity.

Images