JP3628550B2 - Sales prediction method and apparatus, and recording medium on which sales prediction program is recorded - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a sales prediction method and apparatus, and a recording medium on which a sales prediction program is recorded. It relates to the medium.
[0002]
[Prior art]
For example, in the distribution industry, mass merchandise stores, etc., it is necessary to order products regularly. Orders for this product are made by forecasting the sales volume of the product, which changes due to weather conditions, day of the week, and other variable factors, but the forecast error in sales volume is directly related to sales opportunity loss and disposal loss. The accuracy of the prediction is important.
[0003]
Conventionally, as a sales forecasting method, a multiple regression analysis performed in consideration of weather information such as weather, temperature, and humidity, day of the week, and other fluctuation factors is known. This multiple regression analysis analyzes past sales results, and reflects, for example, an analysis result such as a decrease in sales volume of about 5% when the weather is raining in sales forecasts.
[0004]
[Problems to be solved by the invention]
However, the multiple regression analysis that has been used as a sales forecasting method has a problem that the error between the sales forecast and the sales performance gradually increases after analyzing the past sales performance and creating a prediction formula. is there. In addition, there is a problem in that the accuracy of sales prediction becomes very poor when a change in orientation or a change in popular products occurs due to some factor.
[0005]
This is because the sales volume forecast formula created using multiple regression analysis is an analysis of the sales performance up to the time the forecast formula was created, and the sales performance after creating the forecast formula is considered. There is no cause. On the other hand, it is conceivable to recreate the sales volume prediction formula for each short period, but there is a problem when considering the cost and time required to create the prediction formula.
[0006]
The present invention has been made in view of the above points, and recorded a sales prediction method and apparatus capable of preventing a decrease in prediction accuracy and capable of long-term sales prediction using the same prediction formula, and a sales prediction program. An object is to provide a recording medium.
[0007]
[Means for Solving the Problems]
To solve the above problems, the present invention is claimed in claim 1, in the sales method for predicting a sales volume of goods by using a computer, the computer, the sales volume of the predetermined period of the commodity to predict the sales volume It is considered that the computer affects the sales quantity on the forecasted sales date, by calculating the moving average value by averaging the actual values of the actual value , correcting the moving average value with the partial regression coefficient for the moving average value Calculating a fluctuation forecast quantity of the sales quantity of the product corrected by the partial regression coefficient for the fluctuation factor from the fluctuation factor, and the computer based on the corrected fluctuation forecast quantity of the corrected moving average value The method further includes a step of correcting and calculating the first sales forecast quantity.
[0008]
In this way, the moving average value is calculated by averaging the actual value of the sales quantity of the product for which the sales quantity is predicted for a predetermined period, and the moving average value is corrected with the partial regression coefficient for the moving average value. Calculating a fluctuation forecast quantity of the sales quantity of the product corrected by a partial regression coefficient for the fluctuation factor from a fluctuation factor that is considered to affect the sales quantity on the forecast date, and calculating the corrected moving average value A step of calculating the first sales forecast quantity by further correcting based on the corrected fluctuation forecast quantity, and reflecting the actual value of the sales quantity in the predetermined period immediately before the sales forecast date in the sales forecast. It is possible to improve the follow-up of the predicted value with respect to changes in sales and changes in popular products.
[0009]
Therefore, by predicting the sales quantity based on the sales performance in the predetermined period immediately before the sales forecast date and the fluctuation forecast quantity based on the fluctuation factors of the sales forecast date, it is possible to prevent the accuracy of the sales forecast from being reduced. Sales forecasts can be made in the long term using forecast formulas.
[0011]
Accordingly, it is possible to prevent a significant decrease in the accuracy of the sales prediction, and it is possible to predict the sales in the long term using the same prediction formula.
[0012]
In the sales prediction device for predicting the sales volume of a product, the present invention described in claim 2 affects the actual sales value of the product for which the sales volume is predicted for a predetermined period and the sales volume on the sales prediction date. A first storage means for storing possible variation factors; and a moving average value is calculated by averaging the actual value of the sales quantity for the predetermined period, and the moving average value is corrected with a partial regression coefficient for the moving average value; , Calculating a fluctuation forecast quantity of the sales quantity of the product corrected with the partial regression coefficient for the fluctuation factor, and further correcting the corrected moving average value based on the corrected fluctuation forecast quantity to obtain a first sales forecast quantity It has the 1st arithmetic processing means which computes.
[0013]
Thus, by having the first storage means and the first arithmetic processing means, the actual value of the sales quantity in the predetermined period immediately before the sales forecast date can be reflected in the sales forecast. It is possible to improve the follow-up performance of the predicted value with respect to changes in the value. Therefore, it is possible to prevent a decline in the accuracy of the sales forecast by forecasting the sales quantity with the actual value of the sales quantity for the predetermined period immediately before the sales forecast date and the fluctuation forecast quantity based on the fluctuation factors of the sales forecast date. Sales can be predicted in the long term using the same prediction formula.
[0016]
Further, the invention described in claim 6, the computer, the moving average value to calculate the actual value of the sales volume of a predetermined time period of the product to predict sales volume on average, a moving average value for the moving average value thereof procedure for calculating and procedures to be corrected by the partial regression coefficient, from changes in factors that are believed to affect the sales volume of sales forecasting date, the fluctuation prediction quantity of the sales volume of the commodity which has been corrected in the partial regression coefficients for the variables When records the program for executing the steps of further calculating a first sales forecast quantity is corrected based on the moving average value the correction to the corrected change prediction quantity.
[0018]
By using the recording medium according to the third aspect , the inventions according to the first and second aspects can be realized.
The recording medium for recording this program is a magnetic recording medium for recording information magnetically, such as a CD-ROM, a flexible disk, a magneto-optical disk (MO), etc., a ROM, a flash memory, etc. Various types of recording media such as a semiconductor memory for recording data can be used.
[0019]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 shows a configuration diagram of an embodiment of a computer system for realizing a sales prediction apparatus of the present invention. In FIG. 1, this computer system includes an input device 20, a display device 30, a drive device 40, a recording medium 50, an auxiliary storage device 60, a memory device 70, which are mutually connected by a bus B. The configuration includes an arithmetic processing device 80 and a printing device 90.
[0020]
The input device 20 includes a keyboard and a mouse that are operated by a user of the computer system, and is used to input various operation signals to the computer system. The display device 30 displays various windows and data necessary for operating the computer system. The printing apparatus 90 outputs data such as processing results.
[0021]
A sales prediction program related to the sales prediction device is provided by a recording medium 50 such as a CD-ROM, for example. The recording medium 50 on which the program is recorded is set in the drive device 40, and the program is installed from the recording medium 50 to the auxiliary storage device 60 via the drive device 40.
The auxiliary storage device 60 stores the installed sales prediction program and also stores necessary files, data, and the like. The memory device 70 reads and stores the sales prediction program from the auxiliary storage device 60 when the computer system is activated. The arithmetic processing device 80 executes processing related to the sales prediction device in accordance with the sales prediction program read and stored in the memory device 70.
[0022]
For example, a program installed in the auxiliary storage device 60 is read into the memory device 70 and executed by the arithmetic processing device 80 in the processing of the sales prediction device of the present invention.
Next, the principle of the present invention will be briefly described. FIG. 2 is a diagram illustrating an example of the relationship between the sales quantity and the fluctuation factors such as weather information. By performing the multiple regression analysis using the formula of FIG. 2, a relational expression connecting the result and the factor can be calculated.
[0023]
The target variable Y is determined by several variable factors, and indicates, for example, the sales quantity. The explanatory variables x1 to xn represent the fluctuation factors, and represent, for example, data obtained by quantifying precipitation, temperature, humidity, and the like. The partial regression coefficients β1 to βn are numerical values determined for each explanatory variable x1 to xn, and are determined based on past sales results and the fluctuation factors at that time. The constant Const is a correction term.
[0024]
In other words, the partial regression coefficient can be calculated by analyzing the relationship between the sales quantity and the fluctuation factors such as weather information based on the past sales performance and the fluctuation factors at that time. However, if all partial regression coefficients are calculated by analyzing past sales performance and the fluctuation factors at that time as shown in FIG. 2, the forecast formula for the sales quantity created by the partial regression coefficient is gradually There is a problem that the error with the sales record increases. In addition, there is a problem in that the accuracy of sales prediction becomes very poor when a change in orientation or a change in popular products occurs due to some factor.
[0025]
Therefore, by reflecting the average value of sales performance (hereinafter referred to as moving average value) in the period immediately before the sales forecast (for example, 7 days) in the sales forecast, We considered a sales forecasting method that improves follow-up and prevents a significant decline in the accuracy of sales forecasting.
FIG. 3 is an explanatory diagram of the first embodiment of the sales prediction method of the present invention. The sales volume prediction formula of FIG. 3 is roughly composed of two terms. The first term is the sales volume determined by the moving average value. By reflecting the sales performance of the period immediately before the sales forecast in the sales forecast, the follow-up of the forecast value to the change in the orientation, the change in the best-selling products, etc. improves.
[0026]
The other term is sales volume determined by weather information such as weather, temperature, humidity, etc., day of the week, and other fluctuation factors, and predicts increase / decrease of sales volume due to fluctuation factors of the sales forecast date. That is, the prediction formula of FIG. 3 determines the prediction value based on the moving average value by one term and the prediction value based on the fluctuation factor of the sales forecast date by the other term.
[0027]
FIG. 4 is an explanatory diagram of an example that specifically represents the sales volume prediction formula of FIG. 3. FIG. 5 shows an example of a partial regression coefficient of the sales volume prediction formula of FIG. The equation of FIG. 4 uses a moving average value for 7 days, day of the week, weather, daily precipitation, factor 1, factor 2, factor 3, factor 4, and a constant as a variable factor.
The partial regression coefficients x (1) to x (9) of each variation factor are calculated for each product based on the past sales performance and the variation factor at that time. The partial regression coefficients x (1) to x (9) indicate that the larger the numerical value, the more easily the sales quantity is affected by the variation factor.
[0028]
Here, x (1) is a partial regression coefficient of a moving average value for 7 days. x (2) is a partial regression coefficient for the day of the week. x (3) is a partial regression coefficient of the weather. x (4) is a partial regression coefficient of daily precipitation. x (5) is a partial regression coefficient of factor 1. x (6) is a partial regression coefficient of factor 2. x (7) is a partial regression coefficient of factor 3. X (8) is a partial regression coefficient of factor 4.
[0029]
For example, by analyzing the past sales performance of the product “Bento” and the fluctuation factors at that time, it is found that the sales volume of the lunch is easily affected by the difference in the day of the week. Therefore, by increasing the partial regression coefficient x (2) of the product “Bento” in FIG.
Referring to FIG. 5, the partial regression coefficient x (2) of the product “Bento” is larger than the partial regression coefficient x (2) of other products, so the difference in the day of the week greatly affects the sales forecast of the lunch box. I understand that. Further, since the partial regression coefficient x (5) of the product “Ice” is larger than the partial regression coefficient x (5) of the other products, it can be understood that the difference of the factor 1 greatly affects the ice sales forecast.
[0030]
Further, x (1) is a partial regression coefficient of a moving average value that is an average value of sales results in a period immediately before the sales prediction (for example, 8 days before to 2 days before). It indicates how much you want to use. For example, the product “Sushi” in FIG. 5 has a large partial regression coefficient x (1) compared to the partial regression coefficient of other products, and the difference in moving average value greatly affects the sales forecast of sushi. I understand.
[0031]
In this way, when performing sales forecasts for products that are susceptible to sales volume differences due to differences in moving averages, the accuracy of sales forecasts should be improved by greatly considering the sales performance of the period immediately before the sales forecast is made. Can do.
On the other hand, the product “Ice” in FIG. 5 has a small partial regression coefficient x (1) compared to the partial regression coefficients of other products, and the difference in moving average value does not significantly affect the sales forecast of ice. I understand. In this way, when forecasting sales of products whose sales volume is not easily affected by differences in moving averages, the sales performance in the period immediately before the sales forecast is not greatly considered, For example, the accuracy of sales prediction can be increased by taking into account the daily maximum temperature and the like.
[0032]
In addition, A contained in the prediction formula of FIG. 4 is a 7-day moving average value of sales performance. B shows the difference in sales volume for each day of the week. C represents the difference in sales volume for each weather. D is a numerical value (for example, 15 mm → 15) of daily precipitation. E is a numerical value of factor 1. F is a numerical value of factor 2. G is a numerical value of factor 3. H is a numerical value of factor 4.
[0033]
FIG. 6 is a diagram showing an example in which the difference in sales quantity for each day of the week and weather is quantified. The numerical values in FIG. 6 are calculated for each product based on the past sales performance and the fluctuation factors at that time. This numerical value indicates that the larger the numerical value, the larger the sales quantity.
For example, the value of Saturday and Sunday of the product “Bento” in FIG. 6 is larger than the value of other days, so it can be seen that the sales volume of the lunches on Saturday and Sunday is large. Moreover, since the numerical value of clear and cloudy of goods "soft cream" is larger than the numerical value of rain and snow, it turns out that the sales quantity of soft cream at the time of fine and cloudy is large.
[0034]
As described above, according to the sales volume prediction formula of FIG. 3, it is possible to adjust the percentage of the moving average value used for sales prediction by adjusting the partial regression coefficient x (1). Therefore, it is possible to improve the follow-up of the predicted value when the change in orientation and the best-selling product change greatly.
Next, a second embodiment of the present invention will be described. FIG. 7 shows an explanatory diagram of a second embodiment of the sales prediction method of the present invention. The sales quantity prediction formula of FIG. 7 uses the error between the predicted value of the past sales quantity and the actual actual value when the sales quantity is predicted next. In FIG. 7, a (1) and a (2) are adjustment coefficients and correction terms calculated using, for example, an error between the predicted value and the actual value for a plurality of times once or a plurality of times.
[0035]
The calculation method of the adjustment coefficient a (1) and the correction term a (2) in FIG. 7 includes a method of using an average value, a method of obtaining a predicted value by a single correlation of actual values, a method of obtaining by a filter theory (Kalman filter), and the like. It is done.
At this time, a (1)> 1 is effective for correction when the sales quantity monotonously increases with time. a (1) <1 is effective for correction when the sales quantity monotonously decreases with time. a (2)> 0 is always effective for correction when a positive error exists. Further, a (2) <0 is effective for correction when a negative error always exists.
[0036]
In general, the conventional formula for predicting the sales volume has a problem that the error between the sales forecast and the sales performance gradually increases. However, this error tends to increase in either the positive direction or the negative direction, but the characteristics of daily sales forecast changes and sales performance changes are often similar.
Therefore, by correcting the calculated forecast value and eliminating the error that is larger in either the positive direction or the negative direction, it is possible to prevent the deterioration of the accuracy of the sales forecast, and the same prediction formula Long-term sales forecasts are possible. The calculation of the adjustment coefficient a (1) and the correction term a (2) in FIG. 7 is not necessarily performed every time, and may be performed every predetermined period (for example, one week) or the error value. It may be performed when the value becomes equal to or greater than a predetermined value.
[0037]
Next, an example of a system using the sales prediction apparatus of the present invention will be described. FIG. 8 shows a block diagram of an embodiment of a system including the sales prediction apparatus of the present invention.
The system of FIG. 8 includes an information providing center 100 that provides weather information such as weather, temperature, and humidity used for predicting the sales volume, and a POS information storage center 200 that stores POS (Point of sales system) information. , A sales prediction device 300 that performs sales prediction based on weather information, sales results, and the like, and a store 400 that sells products.
[0038]
The store 400 collects daily sales results for each product, and supplies the sales results to the POS information storage center 200 as POS information. Then, the POS information storage center 200 supplies the sales results to the sales prediction device 300 as necessary. Further, the information providing center 100 provides the sales prediction device 300 with weather information such as weather, precipitation, and maximum temperature used to predict the sales volume.
[0039]
The sales prediction apparatus 300 performs sales prediction by the above-described sales prediction method of the present invention using the supplied weather information, sales results, and other information on the sales volume fluctuation factors. Note that the sales prediction device 300 can also perform sales prediction for each store and sales prediction for each region by supplying the sales results for each store 400 from the POS information storage center 200. Then, the sales prediction device 300 supplies the analyzed sales prediction, weather information, and aggregated sales results to each store 400.
[0040]
Therefore, each store 400 can place an order for products based on the sales prediction with high prediction accuracy of the present invention, and sales opportunity loss and disposal loss can be reduced. Further, by providing weather information to each store 400, it becomes possible to grasp the influence of weather information such as weather, temperature, and humidity, day of the week, and other fluctuation factors on the sales quantity for each store 400.
[0041]
The error between the sales prediction and the sales performance is used for a new sales prediction, and it is possible to prevent a decrease in prediction accuracy. In FIG. 8, the sales prediction device 300 is installed at a location different from the store 400, but the present invention is not limited to this, and can be installed for each store 400.
The first storage means described in the claims corresponds to the auxiliary storage device 60 or the memory device 70, the second storage means corresponds to the auxiliary storage device 60 or the memory device 70, and the first arithmetic processing means is an arithmetic unit. Corresponding to the processing device 80, the second arithmetic processing means corresponds to the arithmetic processing device 80.
[0042]
【The invention's effect】
As described above, according to the first aspect of the present invention, the stage of calculating the moving average value by averaging the actual value of the sales quantity of the product whose sales quantity is predicted for a predetermined period, and the sales quantity on the sales forecast date Calculating a fluctuation forecast quantity of the sales quantity of the product from a fluctuation factor that is considered to have an effect; and calculating a first sales forecast quantity by correcting the moving average value based on the fluctuation forecast quantity; By having the actual value of the sales quantity in the predetermined period immediately before the sales forecast date can be reflected in the sales forecast, the follow-up of the forecast value to the change in the orientation, the change in the best selling product, etc. can be improved. .
[0043]
Therefore, by predicting the sales quantity based on the sales performance in the predetermined period immediately before the sales forecast date and the fluctuation forecast quantity based on the fluctuation factors of the sales forecast date, it is possible to prevent the accuracy of the sales forecast from being reduced. Sales forecasts can be made in the long term using forecast formulas.
[0045]
According to the second aspect of the present invention, by having the first storage means and the first arithmetic processing means, the actual value of the sales quantity in the predetermined period immediately before the sales forecast date is reflected in the sales forecast. It is possible to improve the follow-up of the predicted value with respect to changes in orientation, changes in popular products, and the like. Therefore, it is possible to prevent a decline in the accuracy of the sales forecast by forecasting the sales quantity with the actual value of the sales quantity for the predetermined period immediately before the sales forecast date and the fluctuation forecast quantity based on the fluctuation factors of the sales forecast date. Sales can be predicted in the long term using the same prediction formula.
[0047]
Further, according to the present invention described in claim 3, the computer, the procedure for calculating the moving average value actual value of the sales volume of a predetermined time period of the product to predict sales volume and average sales volume of sales predicted day procedure from variables that would affect, for calculating the procedure for calculating the fluctuation prediction quantity of sales of the product, the first sales forecast quantity is corrected based on the moving average value to the fluctuation prediction quantities A program for executing and is recorded.
[0049]
By using the recording medium according to the third aspect , the inventions according to the first and second aspects can be realized.
The recording medium for recording this program is a magnetic recording medium for recording information magnetically, such as a CD-ROM, a flexible disk, a magneto-optical disk (MO), etc., a ROM, a flash memory, etc. Various types of recording media such as a semiconductor memory for recording data can be used.
[Brief description of the drawings]
FIG. 1 is a configuration diagram of an embodiment of a computer system that realizes a sales prediction apparatus of the present invention.
FIG. 2 is a diagram illustrating an example of a relationship between a sales quantity and a variation factor such as weather information.
FIG. 3 is an explanatory diagram of a first embodiment of the sales prediction method of the present invention.
FIG. 4 is an explanatory diagram of an example that specifically represents the sales volume prediction formula of FIG. 3;
5 is a diagram illustrating an example of a partial regression coefficient of the sales volume prediction formula of FIG. 4;
FIG. 6 is a diagram illustrating an example in which a difference in sales quantity for each day of the week and weather is quantified.
FIG. 7 is an explanatory diagram of a second embodiment of the sales prediction method of the present invention.
FIG. 8 is a configuration diagram of an embodiment of a system including a sales prediction apparatus according to the present invention.
[Explanation of symbols]
20 Input Device 30 Display Device 40 Drive Device 50 Recording Medium 60 Auxiliary Storage Device 70 Memory Device 80 Arithmetic Processing Device 90 Printing Device 100 Information Providing Center 200 POS Information Storage Center 300 Sales Prediction Device 400 Store

Claims (3)

In a sales forecasting method for forecasting the sales volume of a product using a computer ,
The computer calculates the moving average value by averaging the actual value of the sales quantity of the product for which the sales quantity is predicted for a predetermined period, and corrects the moving average value with a partial regression coefficient for the moving average value;
Calculating a forecasted change in sales quantity of the product, corrected by a partial regression coefficient for the fluctuation factor, from a fluctuation factor that the computer is expected to affect the sales quantity on the forecasted sales date;
The computer further comprises a step of calculating a first sales prediction quantity by further correcting the corrected moving average value based on the corrected fluctuation prediction quantity. In a sales forecasting device that predicts the sales volume of a product,
First storage means for storing the actual value of the sales quantity of the product for which the sales quantity is predicted for a predetermined period, and the fluctuation factors that are considered to affect the sales quantity on the sales forecast date;
The moving average value is calculated by averaging the actual value of the sales quantity for the predetermined period, the moving average value is corrected with the partial regression coefficient for the moving average value, and the correction is performed with the partial regression coefficient for the variation factor A sales prediction unit that calculates a fluctuation predicted quantity of a sales volume of a product and further corrects the corrected moving average value based on the corrected fluctuation predicted quantity to calculate a first sales predicted quantity. apparatus. On the computer,
A procedure to calculate the moving average value by averaging the actual value of the sales quantity of the product whose sales quantity is predicted for a predetermined period, and to correct the moving average value with the partial regression coefficient for the moving average value,
A process for calculating a forecasted fluctuation quantity of the sales quantity of the product, corrected by a partial regression coefficient for the fluctuation factor, from a fluctuation factor that is thought to affect the sales quantity on the forecasted sales date;
A computer-readable recording medium having recorded thereon a program for executing the procedure of further correcting the corrected moving average value based on the corrected fluctuation predicted quantity to calculate a first sales predicted quantity.

Images