JPH03224054A - Sales amount estimating device - Google Patents

Abstract

PURPOSE:To obtain an accurate and minute estimated total amount of sales by setting plural estimated amounts of sales within a knowledge base together with the generating frequencies added to these estimated amounts for each condition of each sales factor. CONSTITUTION:A knowledge base 11 formed on a magnetic recording disk is provided together with an inference engine part 12 which actually carries out an inference, and a user interface device 13 which contains an inference condition input part 13a which inputs various conditions for the inference and an inference result output part 13b which displays and outputs the estimated total amount of sales for each generating frequency, i.e., the inference result. Furthermore the conditions are set within the knowledge base 11 for each factor of sales, and plural estimated sales amounts are set to each of these conditions for each of different generating frequencies. Thus an estimated total amount of sales is obtained by accumulating the estimated sales amounts corresponding to the highest generating frequency of the corresponding conditions for each factor of sales. Then it is possible to obtain a more accurate and minute estimated total amount of sales including an error range.

Description

[Detailed Description of the Invention] [Object of the Invention] (Industrial Application Field) The present invention relates to a sales forecasting device that infers predicted total sales by specifying conditions for each sales factor, and particularly relates to , relates to a sales forecasting device that predicts predicted total sales for each degree of occurrence.

(Prior Art) For example, when opening a store in one location, a market survey of the area (trade area) is conducted in advance to predict the sales if the store is newly opened. Conventionally, a sales forecasting device shown in FIG. 6 has been developed to predict sales. That is, a large number of production rules called l F-THEN rules are stored in the knowledge base 1.
The inference engine unit 2 uses this rule to perform forward inference or backward inference and outputs a conclusion to the user interface device 3. Also. Each production rule stored in the knowledge base 1 can be arbitrarily registered, deleted, or changed by the operator via the knee sign interface 3.

To be more specific, if the newly opened store is a convenience store, the knowledge base 1 includes, for example, the household structure and age structure of the residents in the relevant trade area, and the existence of similar stores, which are considered to have an impact on sales. Conditions are set for each sales factor such as presence/absence information. For example, as shown in Figure 6, the sales factors of household composition are ■
If you are single and have no other stores nearby, and the parenthetical store belongs to a major chain store, each household will spend 300 yen a day on purchases at this store. That is, the predicted sales amount under this condition (2) is 300 yen per day. Similarly, for condition (2) where there are no other stores nearby for a family of two and the store in parentheses is a major chain store, the predicted sales would be 420 yen.

In this way, the predicted sales amount is stored in the knowledge base 1 for each condition of each sales factor, and when each condition is selected and specified for each sales factor from the user interface device 3, the inference engine section 2 is activated. Then, by accumulating the predicted sales under the specified conditions of each sales factor stored in the knowledge base 1, the predicted total sales for one day at the corresponding store is calculated.

Then, the calculation result, that is, the inference result, is printed out on the user interface device 3, or the C
Display output on the RT display section.

However, the sales prediction device that uses this method to predict the expected total sales for a new store also has the following problem.

In other words, this predicted total sales is the sum of the predicted sales that meet each condition for each sales factor, so if the predicted sales for each condition are not very accurate values, they will be accumulated to become one. The reliability of the predicted total sales as a single value also decreases. However, as shown in FIG. 6, although the predicted sales amount corresponding to each condition is expressed as a single numerical value, in reality, the probability of a value around this value may be quite high. On the other hand, there are cases where it is represented by just this value.

Therefore, although the true total sales have the highest probability of being the ultimately calculated predicted total sales, there is also a considerable probability that the true total sales will be around that. However, how likely is it that the true total sales will match the ultimately calculated predicted total sales, or how likely is it that the true total sales will match the predicted total sales? It was not possible to understand what was above the sales amount, or to what extent the true total sales amount was below the predicted total sales amount.

(Problem to be Solved by the Invention) As described above, according to the conventional sales forecasting device, the final predicted total sales are only one numerical value.
To what extent is the predicted total sales accurate?
In addition, since there was no information on the degree of occurrence of upward movement or, conversely, the degree of occurrence of downward movement, it was not possible to obtain an accurate predicted total sales including the error range. There wasn't.

The present invention was made in view of the above circumstances, and by setting a plurality of predicted sales with different occurrence rates for each condition of each sales factor in the knowledge base, it is possible to finally calculate the predicted total sales. can be obtained for each degree of occurrence,
It is an object of the present invention to provide a sales forecasting device capable of obtaining more accurate and detailed predicted total sales including error ranges.

[Structure of the Invention] (Means for Solving the Problems) In order to solve the above problems, according to the sales forecasting device of the present invention, a plurality of predicted sales with different degrees of occurrence assigned to each condition in each sales factor are provided. When combining the knowledge base that stores the high value, the inference condition input section that inputs the conditions selected for each sales factor, and the conditions for each sales factor input from the inference condition input section using the knowledge base The system is equipped with an inference engine section that calculates the predicted total sales for each occurrence degree, and an inference result output section that outputs the predicted total sales for each occurrence degree calculated by the inference engine section. .

(Function) With the sales forecasting device configured in this way, each condition is set for each sales factor in the knowledge base,
Further, for each condition, a plurality of predicted sales amounts are set for different degrees of occurrence. Therefore, by accumulating the predicted sales corresponding to the maximum value of the occurrence degree of the corresponding condition for each sales factor, the predicted total sales can be obtained which is represented by a single numerical value, which is the same as in the conventional device.

Furthermore, in this invention, not only the predicted sales corresponding to the maximum occurrence degree of the relevant condition but also the predicted sales of other occurrence degrees are accumulated for each occurrence degree to calculate the predicted total sales. Therefore, the degree of occurrence of each calculated predicted total sales amount is determined. In this way, the predicted total sales amount is calculated for each degree of occurrence, so the degree of occurrence of each predicted total sales amount can be grasped.

(Example) An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing a schematic configuration of a sales forecasting device according to an embodiment. This sales forecasting device is composed of a computer control program, and includes a knowledge base 11 formed on a magnetic recording disk, an inference engine section 12 that actually executes inference, and various conditions for inference. The user interface device 13 includes an inference condition input section 13a for inputting the inference conditions, and an inference result output section 13b for displaying and outputting the predicted total sales for each occurrence degree, which is the inference result, on a CRT display section, for example.

In the knowledge base 11, tables showing three types of sales factors are stored, as shown in FIGS. 2, 3, and 4, assuming that a new convenience store is being opened, for example. .

FIG. 2 is a table showing the sales factors of the household structure of the residents in the commercial area to which this store belongs, and in this figure, four conditions (■-■) are set. Under the condition (■) where a single person has no other stores nearby and the store in parentheses belongs to a major chain store, the probability of each household purchasing 280 yen per day at this store is 0.4, and the probability of purchasing 290 yen at this store per household is 0.4. The probability of purchasing 300 yen at this store is 0.6, the probability of purchasing 310 yen at this store is 0.75, and the probability of purchasing 320 yen at this store is 0.75. The probability of purchasing at a store is 0.7. That is,
For one condition (2), five types of predicted sales of 280 to 320 yen whose degree of occurrence varies by 0.4 to 0.8 are stored.

Similarly, for the condition ■ that a family of two has no other stores nearby and belongs to this store or a major chain store, there are five types of predicted sales 400 that vary from 0.4 to 0.8.
~440 yen is stored.

FIG. 3 is a table showing sales factors of age composition, and in this figure, four conditions (■-■) are set. If the person is under 20 years old, there are no other stores nearby, and the store is part of a major chain store, then each person will be charged 30 yen per day.
The probability of purchasing yen at this store is 0.6, which is 50
The probability of purchasing yen at this store is also 0.6. Five types of predicted sales of 30 to 50 yen are stored for this condition (2), the degree of occurrence of which changes from 0.6 to 0.8.

Further, FIG. 4 is a table showing sales factors of competitiveness, and in this figure, three conditions (■-■) are set. That is, under condition (3) where there is no strong competing store within a predetermined distance range of this store, the daily sales amount is multiplied by 1.2.

In addition, under condition (2), which exists slightly, the daily sales amount is multiplied by 1.0. Furthermore, if there are many competing stores, the daily sales will be multiplied by 0.8. Note that the degree of occurrence is not set for each of the sales factor conditions (■) to (■).

In this way, a plurality of conditions are set for each sales factor, and by using the knowledge base 11 in which a plurality of predicted sales figures are stored according to the degree of occurrence of each condition, for example, the conditions shown below can be set on a user interface. Inference condition input section 1 of device 13
The processing procedure of the inference engine unit 12 when input from 3a will be explained.

Singles 100 households Under 20 years old 200 people, 2-person family 200 households 20-30 years old 600 people, 3-person family 80 households 80-50 years old 500 people, 4-person family
150 households, 50 years of age or older, 100 people, there are no stores around them.
As a leading chain store, first, calculate each predicted subtotal sales by multiplying the number of households and number of people by each predicted sales under each of the sales factors of number of households and age composition in Figures 2 and 3. is determined for each degree of occurrence. The results are shown below.

(1> Sales factor conditions for household composition ■ Forecast subtotal Occurrence degree 28000 0.4 29000 0.8 30000 0.8 31000 0.75 32000 0.7 Condition ■ Forecast subtotal Occurrence degree 80000 0.4 82000 0.6 84000 0 .8 86000 0.75 88000 0.7 Condition ■ Predicted subtotal Occurrence degree 14400 0.6 15200 0.7 16000 0.8 16800 0.7 171300 0.6 Condition ■ Predicted subtotal Occurrence degree 38000 o17 37500 0.75 39000 0 .8 40500 0.75 42000 0.7 (2) Sales factor condition of age structure ■ Forecast subtotal Occurrence degree 6000 0.6 7000 0.7 8000 0.8 9000 0.7 10000 0.6 Condition ■ Forecast subtotal Occurrence degree 24000 0.5 27000 0.6 30000 0.8 33000 0.75 36000 0.7 Condition ■ Predicted subtotal Occurrence degree 12500 0.7 15000 0.75 17500 0.8 20000 0.75 22500 0.7 Condition ■ Predicted subtotal Occurrence degree 1000 0.4 1500 0.6 2000 0.8 2500 0.75 3H (10,7 The predicted total sales is calculated by adding up all the predicted subtotal sales under each condition described above, and is shown in the competitiveness condition ■. You can get it by multiplying it by 1 or 2.

Conventionally, there was only one expected subtotal sales per condition, so it was sufficient to simply add them, but in this embodiment, there are five expected subtotal sales for each condition. That is, in this example, the five predicted subtotal sales for each condition in each sales factor are added separately to obtain a total of five predicted total sales 01 to G5.
Calculate. In this case, since the degree of occurrence varies somewhat depending on each condition, an average value is taken at the time of addition.

For example, if each predicted subtotal sales amount for condition (■) and each predicted subtotal sales amount for condition (2) in the household composition are added, each added value and each occurrence degree will be as follows.

Condition ■ Condition ■ Additional value Occurrence degree 28000
+14400-42400 (0.4+0.6)/2
-0.529000 +15200-44200 (
0.6+0.7)/2-0.8530000 +1.5
000-45000 (0,8+O,l1l)/2-
0.831000 +18800- 47800 (
0,75+0.7)/2-0.72532[100++
7600-49600 (L7+0.6)/2-0.
65 In this way, the five predicted subtotal sales are added separately for each condition, and finally multiplied by the aforementioned 1 and 2 to finally arrive at the five predicted total sales G.

~Calculate G5.

Predicted total sales Occurrence rate G + -201900x 1.2 = 242280
0.5375G 2 - 214200X 1
．． 2-257040 0. [1625
G 3-226500x 1.2 271800
0.8G 4-238800x 1.2 fog 28B5
80 0.7375G 5-251100X 1.
2 Sue 301325 0.675 5 predicted total sales G, ~G with different degrees of occurrence calculated in this way
, to the inference result output unit 1 of the user interface device 13
3b, and the relationship between the predicted total sales and the occurrence degree is displayed in a graph on, for example, a CRT display section, as shown in FIG.

With the sales forecasting device configured in this manner, as shown in FIG. 5, it is possible to grasp the predicted total sales when opening a new store as a function of the degree of occurrence.

In this case, the predicted total sales amount 03 whose occurrence degree is the maximum value (0, 8) is one predicted total sales amount calculated by the conventional device.

In other words, in the conventional device, the predicted total sales were only 286,580 yen for G3, but the actual total sales were higher than this predicted total sales G. can become.

In this way, not only one predicted total sales G, but also multiple predicted total sales G l +G2, G with different degrees of occurrence
4. By understanding G5, it is possible to understand the rate of occurrence of each predicted total sales amount. Therefore, the predicted total sales can be estimated more precisely, including the error range.

Note that the present invention is not limited to the embodiments described above. In the embodiment, the predicted sales of a convenience store have been described, but the inference target is not particularly limited to convenience stores. In short, each predicted sales amount corresponding to each condition in each sales factor may be shown with different occurrence rates.

Effects of the Invention As explained above, according to the sales forecasting device of the present invention,
A plurality of predicted sales are set in the knowledge base with different degrees of occurrence for each condition of each sales factor. Therefore, since the final "f-measured total sales amount can be obtained for each degree of occurrence, a more accurate and more detailed predicted total sales amount including the error range can be obtained."

[Brief explanation of drawings]

FIG. 1 is a block diagram showing a schematic configuration of a sales forecasting device according to an embodiment of the present invention; FIGS. 2 to 4 are diagrams showing each sales factor table formed in the knowledge base; 5
The figure shows the relationship between the calculated predicted total sales and the degree of occurrence, and FIG. 6 is a block diagram showing the schematic configuration of a conventional sales forecasting device. 11...Knowledge base, 12...Inference engine section, 1
3... User interface device, 13a... Inference condition human power unit, 13b... Inference result output unit.

Claims (1)

[Claims] a knowledge base that stores a plurality of predicted sales with different degrees of occurrence assigned to each condition in each sales factor; an inference condition input unit that inputs conditions selected for each of the sales factors;
an inference engine section that uses the knowledge base to calculate the predicted total sales for each occurrence degree when combining the conditions for each sales factor input from the inference condition input section; A sales forecasting device comprising an inference result output unit that outputs predicted total sales for each calculated occurrence degree.