JPH02299059A - Order forecasting system - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed Description of the Invention] [Industrial Field of Application] The present invention is applicable to short-delivery orders that predict tomorrow's order quantity on a daily basis, such as same-day order acceptance and shipment, and same-day order acceptance and next-day shipment. In business, in particular, it relates to an order prediction system that utilizes past order information to predict the amount of orders and use it for business.

[Prior Art] Conventionally, in the field of prediction processing targeted by the present invention, as described in, for example, Japanese Unexamined Patent Application Publication No. 59-774, original data is divided into time intervals to calculate period index values. Time-series prediction was used to estimate according to the period index value.

[Problem to be solved by the invention] In the above conventional technology.

(1) There are multiple variables such as weather, day of the week, week, season, etc.

(2) There are many irregular disturbances such as special sales.

These points are not taken into account, and simple time-series forecasting has problems in its application to order forecasting.

In the present invention, considering that the original data consists of multiple factors including time factors, we first break down the order amount by factor, remove dispersion factors such as special sales, and then synthesize the order amount for each factor. It is a prediction.

An object of the present invention is to provide an order forecasting system that calculates index values for each variable factor and makes it easier to apply time-series forecasting using an order model that removes irregularities from original data.

Furthermore, another object of the present invention is to provide an order prediction system that makes it easier to judge and correct predictions by providing estimated order information obtained by selecting indexes for each factor and multiplying them by predicted values.

[Means for solving the problem] The above purpose is to create a daily order table with order amounts by day and product for several months, a total sales table with monthly sales totals for several years, and a table by week, day of the week, and a variation removal unit that has a calendar table with weather categories for each day and removes variations from the nine-day order table and creates a general order file;
An index calculation unit that calculates the monthly index from the "Total Sales" table and indexes by week, day of the week, and weather from the general order table to create each index table, and an order model file that removes factors using each index from the general order table. A model creation section that creates orders, an order forecasting section that forecasts in chronological order based on the order forecast period and order model entered from the keyboard, and an estimated order amount edited by day, product, and weather from the order forecast data and factor-specific indexes. This is achieved by an order prediction system that has a weather-specific editing section that outputs to a display screen and printer.

[Effect] First, after removing variations such as special sales from the original data, calculate the monthly average value, divide this by the annual average to calculate the index for each month, and use the index to calculate the monthly factor from the original data. Remove. Next, a monthly average value is calculated from the data from which the monthly factor has been removed and divided by the all-week average value to calculate a weekly index, and the weekly factor is also removed. Similarly, indexes are calculated for the day of the week and weather, and each factor characteristic is removed.

The above calculations are performed for each product to create an order model, and the exponential smoothing method is used to make predictions in time series at a predetermined period.

Next, the estimated order amount is synthesized by multiplying the predicted value by the index for each factor that has already been calculated, and the order amount is estimated for each weather condition: clear, cloudy, light rain, and rain using the weather index.

[Example] Hereinafter, an example of the present invention will be described in detail based on the drawings.

FIG. 1 is a system configuration diagram of the order prediction system, which includes a processing device 1, a daily order table 2 which is a table of original data and processing conditions, a total sales table 3, a calendar table 4. General order file 5, monthly index file 6, week index file 7, day of the week index file 8, weather index file 9, order model file 10, order forecast file 11. Estimated order file 1
2. Display screen 13 for key-in and display printing; keyboard 14; printer 15; variation removal unit 101, which is a function of processing tables and files within the processing device;
Index calculation unit 102, model creation unit 103, order prediction unit 1
04, a factor synthesis section 105, and a weather editing section 106.

First, the basic flow of the present invention will be explained with reference to FIG.

First, as a preparatory process, create a general order file 5 that removes variations from the daily order table 2 (101).
, Calculate the month, week, day of the week, and weather index from the total sales table 3, general order file 5, and calendar table 4.10
2), monthly index file 6 and weekly index file 7, respectively.
, day of the week index file 8, and weather index file 9, and an order model file 10 is created by removing each index from the general order file 5 (103).

Next, the forecast period is entered manually using the display device 13 and the keyboard 14, and the order volume within the target period is calculated for each product by day using the exponential smoothing method using the order model data to create an order forecast file 11. (t04).

Next, using the monthly index file 6, weekly index file 7, and day of the week index file 8 that have already been created for each factor, the order forecast data is multiplied by factor and the order amount is synthesized, and the estimated order file 1
2 (105).

Finally, receive the weather index from weather index file 9,
The estimated order amount is multiplied by the estimated order file 12 and the estimated order amount by day, product and weather is printed out on the printer 15 (10
6).

Next, the processing contents of the order prediction system shown in FIGS. 1 and 2 will be explained with reference to FIG. 3.

In Figure 3, 301, 302, 303, 304.3
05 corresponds to the preparatory process in FIG. 2, 306.307,
308 corresponds to the processing of order prediction 104, factor synthesis 105, and weather editing 106 in FIG.

First, calculate the average sales for each month and index it (30
1) Using this seasonal index, a model is created in which seasonal factors are removed from daily sales information (302). Next, we remove the week factor, day of the week factor, and weather factor from this model.
03,304,305).

We used an analytical model that excluded each of the factors mentioned above.

Past data is analyzed and trend values are captured (306).

Then, the previously calculated index (season, week, day of the week) is added to the trend value (307). As shown in the table below, use a quick reference table that summarizes trends by weather to predict tomorrow's production volume by d1 (30.8).

Next, in order to remove seasonal factors, weekly factors, day of the week factors, and weather factors, the method of indexing each factor is shown in Figure 4 (
This will be explained in A), (B), (C), and (D).

FIG. 4(A) explains the method of calculating the seasonal index and the method of removing seasonal factors.

(a) First, calculate the monthly average sales and annual average sales for the past three years, and then calculate the monthly seasonal index using the formula shown in (b). Next, according to the procedure shown in (c), the daily sales data is smoothed using the calculated seasonal index, and this is used as model (2).

Further, according to FIG. 4(B), a method for calculating the week index and a method for removing the week factor will be explained.

(a) First, using the data of model (2) from which seasonal factors have been removed, weekly indexes are calculated by the procedure shown in (b).

Next, model (2), which is sales data obtained by smoothing the weekly index, is calculated by the procedure shown in (C).

Therefore, model ■ is sales data with seasonal and weekly factors removed.

Next, according to steps (a), (b), and (c) shown in Figure 4 (C), the day of the week index is calculated and the day of the week factor is removed, and the sales data from which seasonal factors, weekly factors, and day of the week factors are removed is Calculate a certain model ■.

Finally, calculate the weather index and remove weather factors by following steps (a), (b), and (C) shown in Figure 4 (D), and sales after removing seasonal factors, week factors, day of the week factors, and weather factors. Calculate an analytical model with high data.

The data of this analytical model enables the process 306 in FIG. 3 to predict the order amount for each product using regression analysis or other appropriate time-series analysis techniques.

This corresponds to process 104 in FIG.

Next, at 307 in FIG. 3, the predicted order amount is multiplied by the calculated seasonal index, week index, and day of the week index to predict the order amount taking seasonal factors, week factors, and day of the week factors into consideration. this is,
This corresponds to the process 105 in FIG.

Next, at 308 in FIG. 3, the predicted order amount is multiplied by the calculated weather index, and the order amount that takes seasonal factors, weekly factors, and day of the week factors into consideration is output as an estimated order amount that is subdivided by weather. This corresponds to process 106 in FIG. Therefore,
The person in charge of forecasting can refer to the weather-specific estimated orders Ji15 and obtain the predicted order amount corresponding to the weather on that day.

Note that examples of operating procedures according to the processing described above are shown in FIGS. 5(A), (B), (C), and (D). That is, (1)
Calculation of the number of each moon phase, (2) deletion of seasonal factors, (3) deletion of week factors, (4) deletion of day of the week factors, (5) deletion of weather factors.

(6) Creating an analytical model; (7) Processing next month's order prediction.

[Effects of the Invention] The present invention provides an order prediction system that utilizes past order information to predict the order amount and use it for business purposes in industries that require short delivery orders and produce by predicting tomorrow's order amount on a daily basis. This has the effect that it can be obtained.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing an embodiment of the present invention, Fig. 2 is a diagram showing the concept of this system, Fig. 3 is a flowchart showing the prediction procedure, and Fig. 4 is a diagram for explaining the details of the prediction procedure. FIG. 5 is a diagram for explaining the operating procedure. Representative Patent Attorney Katsuo Ogawa Figure 2 Figure 3 Figure 4 (A) (a) Calculate the monthly average sales for the past three years. <
(e, sales) (b) Calculation of seasonal index (C) Smooth sales information by region, product, and day for the past three months using the seasonal index. Figure 4 (B) (b) Calculation of weekly index Average sales for the first week tfiIMLMf1 Loss = Monthly average land filling amount ÷ Number of business days 8100 (C) Smooth the daily sales model ■ using the weekly index FIG. 4(C) (a) Calculate the average sales for each day of the week from the daily sales model (■) smoothed by the weekly index. (b) Calculating the day of the week index (C) Smoothing the daily IH high mo model using the day of the week index. Figure 4 (D) (a) Classify the daily sales model ■ into clear, cloudy, light rain, and rain according to the probability of rain, and calculate the average sales by weather. Total sales on sunny days Average sales on sunny days = Sunny days. Number of days (b) Calculation of weather index Average sales on sunny days - Average sales on days (C) Daily sales model 2 is smoothed by the weather index. 1315 Figure (A) Figure 5 (B) Figure 5 (C)! Figure 15 (D)

Claims (1)

[Scope of Claims] 1. In a system that calculates the future order amount for each product from past original data that fluctuates almost periodically, consisting of dates, products, and order amounts from which variations have been removed, An order forecasting system characterized by setting an order model from which fluctuation factors have been removed, and synthesizing a predicted order amount with an index for each factor into the order model. 2. It is characterized by setting an order model that classifies the weather using the probability of rain and making it all sunny, and by combining the order model and the order amount according to the weather category to remove the weather factor from the original data. Order forecasting system.