JP2005078277A - Shipping quantity predicting system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a system for predicting the practical shipping prediction quantity of merchandise. <P>SOLUTION: This shipping quantity prediction system is provided with an information input means 5 for inputting the past shipping quantity result/weather conditions, a data system preparation means 6 for setting the data system of dependent variables and the data system of explanation variables from the information of the information input means 5, a multiple regression analysis means 7 for generating a multiple regression model adapted to the dependent variable data system and the explanation variable data system, and for outputting a shipping quantity result logical value and a shipping quantity prediction value by using the multiple regression model and a correction means 8 for setting the critical value of the shipping result value as a correction shipping quantity result value, and for calculating the rate of the correction shipping quantity result value to the shipping quantity result logical value as a correction value, and for correcting the shipping quantity prediction value by the correction value. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention is a container-packed beverage product whose shipping amount is likely to fluctuate depending on weather conditions, date, day of the week, business day, etc., for example, a product in which beverages such as soft drinks, beer, sparkling liquor, Chu-hi are packed in bottles, cans or paper packs The present invention relates to a shipping amount prediction system for predicting a shipping amount that is an effective index for determining the production amount.

  Here, the system means hardware controlled by software.

  As an apparatus for predicting the sales of products and the predicted amount of shipment based on past results, there are technologies disclosed in Japanese Patent Application Laid-Open Nos. 4-171777 and 8-212191. A method of multiple regression analysis for explaining a dependent variable based on a plurality of explanatory variables is conventionally known.

Japanese Patent Application Laid-Open No. 8-212191 discloses a method of selecting explanatory variables that well explain dependent variables and predicting sales volume from these explanatory variables by multiple regression analysis.
Japanese Patent Laid-Open No. 4-171577 JP-A-8-212191 C. Hayasi (1952): On the prediction of phenomena from qualitative data from the mathematico-statistical point of view. Annals of the Institute of Statistical Mathematics, 3,69-98. Shigenobu Aoki (1989): Medical Statistical Analysis Reference Manual, Medical School (Tokyo)

  The conventional multiple regression analysis technique is to discover and select an explanatory variable that explains the dependent variable, or to discover and generate a multiple regression equation (multiple regression model) that fits well with the selected explanatory variable and dependent variable. The main focus was placed.

  However, in general, the past sales performance is used as one of the explanatory variables, but the past sales performance inevitably includes the influence of temporary fluctuation factors that affect the sales performance. For example, holidays, events, sales activities, etc. are representative examples of the temporary fluctuation factors.

  For this reason, it has been impossible to obtain a predicted amount of shipment of a practical product only by data processing based on conventional multiple regression analysis.

  On the other hand, it is possible to identify the temporary fluctuation factors and eliminate the influence from the multiple regression equation, but it is difficult to specify all of the temporary fluctuation factors, and even if they are specified, each fluctuation It is difficult to assess the magnitude of the influence of factors.

  Therefore, the problem to be solved by the present invention is to obtain a practical estimated value of the shipping amount, using the actual shipping amount as a dependent variable, and the temporary variable factors such as weather, business days, and colden week as explanatory variables. And providing a practical shipping amount prediction system by appropriately correcting the values of the dependent variables (shipment amount prediction values, etc.) obtained by the multiple regression equation.

  A shipping amount prediction system for container-packed beverage products according to the present invention includes information input means for inputting past shipment volume results and past weather condition information, and a past predetermined period from the information input by the information input means. Information on past weather conditions in which the information input means inputs a data series of explanatory variables corresponding to the data series of the dependent variable, setting the actual quantity of shipment and the quantity of shipment in the period to be predicted as a data series of the dependent variable A data series creating means for acquiring and setting from the above, a multiple regression model adapted to the dependent variable data series of the past shipment volume results and the explanatory variable data series of the weather conditions corresponding thereto, and generating the multiple regression model Output the actual shipment value theoretical value from the explanatory variable data series of the weather conditions corresponding to the dependent variable data series of the past shipment results, and Multiple regression analysis means for outputting a predicted shipment amount from the data series of explanatory variables of the weather conditions corresponding to the dependent variable data sequence of the shipment amount for the period to be attempted, the actual shipment amount actual value and the corresponding shipment amount The actual value and the predicted shipment amount are input, the rejection value of the actual shipment amount is used as the corrected shipment actual value, and the ratio of the corrected shipment actual value and the actual shipment amount theoretical value is calculated as the correction value. And a correction means for correcting the predicted shipping amount with the correction value and outputting the corrected predicted shipping amount.

  According to the present invention, a multiple regression equation is obtained from a past weather condition and a past actual use product by multiple regression analysis means, and a past weather condition corresponding to a period to be predicted using the multiple regression equation. The predicted shipment amount can be calculated from Furthermore, in order to eliminate the effects of temporary fluctuation factors without specifying temporary fluctuation factors, the reject value of the actual shipment volume product value is used as the corrected shipment volume actual value, and the actual shipment volume actual value and the corrected shipment volume actual value. The ratio is calculated as a correction value, and the predicted shipment amount is corrected by this correction value. As a result, it is possible to predict the practical shipment amount without specifying the temporary fluctuation factor and eliminating the influence of the temporary fluctuation factor.

  The information input means inputs information on past weather conditions, business days, and special business days, and the data series creation means includes information on past weather conditions, business days, and special business days input by the information input means. Data is acquired and a data series of explanatory variables corresponding to the data series of the dependent variable is set, and the multiple regression analysis means includes the dependent variable data series of the past actual shipment amount and the explanatory variable data series corresponding thereto. A multiple regression model conforming to the above is generated, and a shipping quantity actual theoretical value is output from the explanatory variable data series corresponding to the dependent variable data series of the past shipment actual using the multiple regression model, and the prediction is made The predicted shipment quantity can be output from the data series of the explanatory variables corresponding to the dependent variable data series of the shipment quantity for the period to be processed.

  According to the present invention, in addition to past weather conditions, relatively obvious temporary fluctuation factors such as business days and business special days are used as explanatory variables, and from these explanatory variables and the actual shipment quantity values as dependent variables, A highly accurate multiple regression equation can be found and generated. Furthermore, in order to eliminate the influence of temporary fluctuation factors such as events, special sales activities, and other factors, the reject value of the actual shipment value is used as the corrected shipment actual value, and the actual shipment value theoretical value and the corrected shipment amount. The ratio with the actual value is calculated as a correction value, and the predicted shipping amount is corrected by this correction value. Thereby, a practical shipment amount prediction can be performed.

  The correction means can make data out of the 95% confidence interval of the actual shipment quantity theoretical value the corrected shipment quantity actual value.

  Here, the weather conditions include temperature (average temperature, maximum / minimum temperature), humidity, fine weather, rainy weather, cloudy weather, and the like.

  The business day information includes information such as date, day of the week, presence / absence of business, number of business days, and the like.

  The special business days are days that affect the shipment volume, such as the beginning of the month, the end of the month, the Bon Festival, before and after the Bon Festival, Golden Week, public holidays, the end of the year, and the predetermined day of the week when the following Monday is a holiday.

  By setting the actual shipment volume as the dependent variable, mainly setting the weather conditions as the explanatory variable, and performing correction to eliminate temporary fluctuation factors from the value of the dependent variable (shipment volume predicted value etc.) obtained by the multiple regression equation, A predicted value of the shipping amount that can be put to practical use can be obtained.

  Embodiments of the present invention will be described below, but the present invention is not limited thereto.

  FIG. 1 shows a configuration of a shipping amount prediction system for container-packed beverage products according to an embodiment of the present invention.

  In FIG. 1, the code | symbol 1 shows the shipment amount prediction system of the container stuffed drink goods by this embodiment.

  The shipping amount prediction system 1 includes a processing unit 2 or a processing unit 2 of a single processing unit, and includes an input / output unit 3 and a storage device or another computer 4 as necessary depending on the system configuration.

  The processing means 2 composed of a processing device group or a single processing device has information input means 5, data series creation means 6, multiple regression analysis means 7, and correction means 8.

  The shipment amount prediction system 1 may be a single computer or a plurality of computers that are communicably connected.

  When the shipment amount prediction system 1 is a single computer, preferably the information input means 5, data series creation means 6, multiple regression analysis means 7 and correction means 8 are processed by a physically identical CPU under program control. Depending on the stage, the information input means 5, data series creation means 6, multiple regression analysis means 7 and correction means 8 operate.

  In that case, the input / output means 3 is preferably an input / output means of the computer. The storage device 4 is a storage device of the computer.

  On the other hand, when the shipment amount prediction system 1 has a plurality of computers communicably connected, the information input means 5, data series creation means 6, multiple regression analysis means 7 and correction means 8 are separate computers that perform data processing. It can be.

  In that case, the input / output means 3 can be a computer that inputs / outputs data and commands to / from each computer that is the information input means 5, the data series creation means 6, the multiple regression analysis means 7, and the correction means 8. . In this case, the input / output means 3 can be a computer external to the shipment quantity prediction system 1, and the input / output means 3 is not necessarily included in the shipment quantity prediction system 1.

  In this case, the storage device or other computer 4 can be a separate storage device or computer that stores and provides information such as weather conditions, business days, special business days, and the like. It is not necessary to include another computer 4 in the shipping amount prediction system 1.

  The information input means 5 is a means for inputting past shipment volume results and past weather condition information from a storage device or another computer 4.

  The data series creation means 6 is a means for setting the data series of dependent variables and the data series of explanatory variables from the information input by the information input means 5. The dependent variable data series is a set of shipment volume data for a predetermined period, and the explanatory variable data series is a set of data such as weather conditions having a predetermined period relationship with the dependent variable data series. That is, the data series creation means 6 prepares the data series of the dependent variables and the data series of explanatory variables so as to have a certain correspondence.

  The multiple regression analysis means 7 generates a multiple regression equation that best fits the correspondence between the explanatory variable data series and the dependent variable data series, and uses the predetermined multiple regression formula to calculate the dependent variable data series from the explanatory variable data series. Is a means for generating

  This is a means for excluding the influence of temporary fluctuation factors from the predicted shipment quantity predicted by the correction means 8 and the multiple regression analysis means 7.

  Hereinafter, specific processing of the information input unit 5, the data series creation unit 6, the multiple regression analysis unit 7, and the correction unit 8 will be described with reference to FIG.

  FIG. 2 is a flowchart showing a processing flow of the shipment amount prediction system 1 according to the present embodiment.

  FIG. 4 schematically shows a flow of forecasting the shipping amount described below. It will be easier to understand by referring to FIG.

  The multiple regression analysis is a method for estimating or predicting the value of an external criterion (dependent variable) quantitatively measured by a linear expression of quantitative data (explanatory variable). Is intended to obtain a practical prediction value by correcting the prediction value obtained by the multiple regression equation by an appropriate method.

  As shown in FIG. 2, when the shipping amount prediction system 1 of the present embodiment starts processing (step S100), the user first selects an explanatory variable to be used as a dependent variable (step S110), and information input means. 5 inputs the data of the explanatory variable and the dependent variable.

  The data series creation means 6 generates and sets a data series of explanatory variables and dependent variables as follows, for example, according to the setting by the user.

  First, with regard to past weekly shipment quantity, the theory of weekly shipment quantity results up to the most recent one week (52 weeks) from the start of product shipment forecasts (the week immediately before the forecast of shipment quantities). The value (A) (total unit from Monday to Sunday) is taken as the dependent variable data series. As for the future weekly shipment quantity, the weekly shipment quantity forecast value (B) for the next 18 weeks from the commodity shipment forecast start week is used as the data series of the dependent variable.

  Next, with regard to explanatory variables, for past product shipments, the weekly shipment volume results for the past one year (52 weeks) from the product shipment forecast start week (total units from Monday to Sunday) As an explanatory variable, the average value of the highest temperature for one week shifted by -4 days (a unit from every Thursday to the current Wednesday) is highly correlated. In addition, for future product shipments, it is shifted by -4 days for the next 18 weeks from the product shipment start week (unit from Thursday of last week to Wednesday of this week). Let the average value be the explanatory variable.

  FIG. 3 shows the relationship between the week having the actual shipment volume and the temperature corresponding week corresponding to the week. Here, the corresponding period is not limited to -4 days, and may be an arbitrary period depending on the consideration of the relationship between the product shipment and the weather conditions.

  Furthermore, since business days and special business days are selected as explanatory variables in this embodiment, the number of weekly business days for the next 18 weeks for which shipping is predicted from the product shipment start week is used as explanatory variables. Further, as a dummy variable, since the shipment of the previous week increases in the week in which Monday starts with a holiday, that week is designated (for example, 1 is set when Monday of the next week is a holiday, and 0 is set otherwise). In addition, as a dummy variable, 1 is set at the beginning of the month, and 0 is set when it is not hit. Furthermore, as a dummy variable, the week in which shipment fluctuates rapidly after the end of the Bon Festival, after Bon Festival, Golden Week, after Golden Week, at the end of the year and the end of the year is set as the shipment quantity prediction system 1, and the other weeks are set as 0 (step in FIG. 2). S120).

  As described above, when the data series creation means 6 sets the data series of the explanatory variable and the dependent variable, it passes the data series of the explanatory variable and the dependent variable to the multiple regression analysis means 7.

  The multiple regression analysis means 7 generates a multiple regression model (multiple regression equation) that fits the dependent variable data series of past shipment volume results and the corresponding explanatory variable data series (step S130).

  In addition, the multiple regression analysis means 7 uses the multiple regression model (multiple regression equation) to calculate the actual shipment quantity actual value (output the actual shipment quantity value) from the explanatory variable data series corresponding to the dependent variable data series of the past shipment results. Output a theoretical value (A) of weekly shipment volume up to the point in time, and a provisional shipment volume forecast value from the data series of explanatory variables corresponding to the dependent variable data series of the shipment volume for the period to be forecasted (Weekly shipment amount forecast value (B) for the next 18 weeks) is output.

  Next, the shipping amount prediction system 1 of the present invention corrects the provisional value of the shipping amount prediction value by the correction means 8.

  First, the correction means 8 estimates the interval by t-test of the actual values of the latest 5 weeks (step S140), then replaces the rejection value with the average value of 5 weeks (step S150), and then the theoretical value of the latest 5 weeks. And an error average (F) of the actual values is calculated (step S160).

Specifically, in order to correct the market trend from the provisional value of the forecasted shipment quantity, the average (C) of the shipment quantity actual value series (D) for the five weeks going back including the shipment quantity actual of the week before the forecast is performed (C ), And then, to eliminate fluctuations such as special sales activities, replace the data that falls outside the 95% confidence interval with the average (C), and create a corrected shipment volume actual value series (D ′) for 5 weeks . Next, the sum (E) of the difference between the corrected shipment quantity actual value series (D ′) for five weeks and the weekly shipment quantity actual theoretical value (A) is set to five weeks of the corrected shipment quantity actual value series (D ′). Divide by the sum of minutes (D ″) to obtain the correction value (F).
here,

  Next, the correction means 8 uses the correction value (F) to correct the provisional value actual shipment amount value to the final shipment amount actual value (corrected shipment amount prediction value) (steps S170 and S180).

  Specifically, the correction means 8 divides the above-mentioned provisional value, the weekly average shipment amount predicted value (B), by the correction value (F) ((B) / (1- (F)) for the next 18 weeks. The corrected weekly shipment amount predicted value (B ′), that is, the final predicted shipment amount (corrected shipment amount predicted value) is calculated.

  The above processing is repeated for the product (item) (step S190 in FIG. 2), and when the processing is completed for all items, the entire processing is terminated (step S200 in FIG. 2).

  In the above description, the actual value and the predicted value of the shipment amount are the weekly shipment amount actual value and the predicted value. However, the actual value and the predicted value are not limited to the week unit, and are not limited to the weekly unit. Can be included).

  The invention of the present application is not limited to the prediction of the shipping amount of beverage products, but can be applied to the use of shipping prediction of products in other fields whose shipping amount depends on weather conditions.

1 is a block diagram showing a configuration of a shipment amount prediction system according to an embodiment of the present invention. The flowchart which showed the flow of the process of the shipment quantity prediction system by one Embodiment of this invention. Explanatory drawing which showed typically the relationship between the week with a shipment actual value and a predicted value, and a temperature corresponding week. Explanatory drawing which represented the flow of the prediction of the shipment quantity typically.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Shipment amount prediction system 2 Processing unit or processing unit of single processing unit 3 Input / output unit 4 Storage unit or other computer 5 Information input unit 6 Data series creation unit 7 Multiple regression analysis unit 8 Correction unit

Claims (3)

Information input means for inputting past shipment volume results, past weather condition information,
Data of the explanatory variables corresponding to the data series of the dependent variable is set as the data series of the dependent variable by setting the actual shipping quantity of the past predetermined period and the shipping quantity of the period to be predicted from the information input by the information input means. Data series creation means for obtaining and setting a series from information on past weather conditions input by the information input means;
Generate a multiple regression model that fits the dependent variable data series of the past shipment volume results and the explanatory variable data series of the weather conditions corresponding thereto, and uses the multiple regression model to determine the dependent variable data series of the past shipment results Shipment from the data series of the explanatory variables of the weather conditions corresponding to the dependent variable data series of the shipment quantity for the period to be predicted Multiple regression analysis means for outputting a quantity prediction value;
The actual shipment amount actual value, the corresponding actual shipment amount value, and the predicted shipment amount are input, the reject value of the actual shipment amount value is set as the corrected actual shipment amount value, and the corrected actual shipment amount value and the shipment A correction means for calculating a ratio of the actual quantity theoretical value as a correction value, correcting the shipping quantity predicted value by the correction value, and outputting a corrected shipping quantity predicted value;
A shipping volume prediction system for container-packed beverage products, comprising: The information input means inputs past weather conditions, business days, special business day information,
The data series creation means obtains data from the past weather conditions, business days, and special business day information input by the information input means, and sets a data series of explanatory variables corresponding to the data series of the dependent variables. ,
The multiple regression analysis means generates a multiple regression model that matches the dependent variable data series of the past shipment volume results and the corresponding explanatory variable data series, and uses the multiple regression model to determine the past shipment results. Shipment volume actual theoretical value is output from the explanatory variable data series corresponding to the dependent variable data series, and the shipping quantity is predicted from the explanatory variable data series corresponding to the dependent variable data series of the shipping quantity in the period to be predicted 2. The shipping amount prediction system for container-packed beverage products according to claim 1, wherein a value is output.   3. The shipping quantity prediction system for packaged beverage products according to claim 1 or 2, wherein the correction means sets data out of a 95% confidence interval of a shipping quantity actual theory value as a corrected shipping quantity actual value.

Images