JPH05189406A - Customer number/sales estimating device applying fuzzy inference - Google Patents

Abstract

PURPOSE:To exactly and easy estimate the customer number and the sales of a restaurant, etc., by applying a computer system and inferring the customer number and the sales from each parameter with use of a membership function and the fuzzy rules. CONSTITUTION:When the customer number and the sales of the next day, for example, an operator inputs the weather and the temperature of the next day through an input/output part 3 based on the weather forecast, etc. Meanwhile a fuzzy inference part 5 selects properly the necessary rules based on the relevant day of the week and time band and carries out a fuzzy inference with a maximum/minimum method, etc. That is, the degree of satisfaction of each rule is obtained at the rule antecedent part based on a membership function and with a head-out method. Then the head of the membership function is cut for the consequent part of each rule based on the obtained degree of satisfaction. Then the function centroids are synthesized together so that the conclusions (customer number and sales) are obtained.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for predicting the number of customers and sales at restaurants and the like, and more particularly to a device for performing prediction calculation by fuzzy inference.

[0002]

2. Description of the Related Art Generally, in restaurants such as family restaurants and taverns, it is important to predict the number of customers and sales because the amount of materials purchased and the number of employees required by time of day depend on the number of customers and sales. If this prediction is greatly deviated, the amount of materials purchased and the number of employees will be insufficient, which will hinder store operation.

The number of customers is affected by factors such as the weather and the day of the day, and factors such as an increase in the number of customers and the unit price of customers at the end of the month. Therefore, it is difficult to accurately predict the number of customers. For this reason, conventionally, in general, it is common to plan the number of customers to purchase the next day, for example, by forecasting the number of customers, etc., by taking into consideration such things as a book (past records), weather forecasts, and other information, and skilled persons such as the person in charge of the store, etc. Target.

[0004]

However, as described above, in the method of predicting in consideration of the number of customers and the like, the prediction accuracy depends on the individual ability, and it is not always possible to expect highly reliable prediction. Moreover, when a forecasting expert is absent on vacation or the like, proper forecasting becomes difficult.
For this reason, it has been a great burden on the predictor.

In view of such circumstances, it is an object of the present invention to introduce a computer system so that the number of customers and sales of restaurants and the like can be accurately and easily predicted.

[0006]

Means and Actions for Solving the Problems
In order to achieve the above object, a device for predicting the number of customers and sales by fuzzy reasoning is provided. Various parameters that affect the number of customers are input to this device. This parameter shall include the weather forecast and regular holidays at a minimum. Further, the knowledge base stores preset membership functions and fuzzy rules. The membership function is set for each parameter and each of the fuzzy labels regarding the number of customers and sales. The fuzzy rules are set with the antecedent part as the parameter and the consequent part as the number of customers and sales. If parameters are input to this device, fuzzy inference is executed using the above knowledge, and the predicted value of the number of customers and sales is calculated. When a predicted value for each time zone is required, a time zone may be added to the above parameters, or a fuzzy rule may be set for each time zone and used properly.

By the way, the knowledge base needs to be constructed according to the situation of the target store. Therefore, the present invention provides a device for supporting the work in constructing the knowledge base. That is, the number of customers and sales are pre-aggregated according to predetermined items such as weather, holidays, etc., and the aggregated result is accumulated as past performance data.

[0008] Then, when constructing the knowledge base, a rough estimate of the number of customers / sales (approximate estimate) using past accumulated data is used.
To calculate. If this result is output by a CRT or the like, it can be used as a reference when the operator creates or modifies the membership function or the like. The rough prediction value is obtained by correcting the data of the same period as the forecast period with the actual growth rate. For the actual growth rate, for example, the data for the last three months is extracted from the past actual data and set as the latest period data,
It is obtained by calculating the increase rate of the latest period data with respect to the data over the past year.

[0009]

Embodiments of the present invention will be described below with reference to the drawings.

1. System Configuration FIG. 1 shows an example of the overall system configuration. Statistical processing device 1
Is for processing book data such as the number of customers and sales, and obtaining past results. The number of customers / sales forecast system 2
It is for reasoning sales. 3 is a keyboard and CR
An input / output unit including T and the like.

In the statistical processing device 1, statistical processing is performed using the book data, and actual result data showing the past actual results are created and accumulated. The method of inputting data related to sales, the type of data, the system configuration, etc. differ depending on the company / store. Normally, it is considered that the items shown in Tables 1 and 2 should be included at a minimum. In other words, it is considered necessary to collect daily data and hourly data for items such as the number of customers, sales, and customer unit price.

[0012]

[Table 1]

[0013]

[Table 2]

A data transfer method from the statistical processing device 1 to the customer number / sales forecasting system 2 can be arbitrarily selected. The customer / sales forecasting system 2 side only needs to support serial connection by RS232C, connection by Ethernet (TCP / IP), and the like.

The function of the customer number / sales forecasting system 2 comprises a knowledge base creation support unit 4 and a fuzzy inference unit 5. The knowledge base creation support unit 4 has a function of processing the result data from the statistical processing device 1 and providing it to the operator as a format suitable for reference when creating the knowledge base. In other words, it has a performance totaling unit 6 that collects performance data to obtain knowledge creation base data, and a knowledge creation base 7 that stores this knowledge creation base data, and data extraction that performs support processing during knowledge creation. It has a section 8, an elongation rate calculation section 9, and a rough prediction section 10. The knowledge base creation unit 11 creates a knowledge base based on the input instruction from the input / output unit 3. On the other hand, the fuzzy inference unit 5 uses the created knowledge to predict the number of customers / sales, and includes a knowledge base 12 and an inference unit 13. Hereinafter, each function will be described in detail.

2. Knowledge creation support function (1) Creation of base data for knowledge creation When creating base data for knowledge creation, first, actual data is aggregated by day of the week. An example of the aggregation process is shown in FIG. As shown in this figure, in this totaling process, the days of the week are classified into three types, weekdays, Saturdays, and Sundays (holidays), and the number of customers and sales are totaled according to the weather for each classification. The aggregated items are average value, maximum value and minimum value. This tabulation is performed on a daily basis and on a predetermined time zone basis. The totaling result thus obtained is stored in the knowledge creation base. According to the output instruction, it is displayed on the CRT in the format of FIG. 2 or printed by the printer. The operator can create a knowledge base with reference to the result of the aggregation.

(2) Knowledge creation support example When creating a knowledge base, for example, 1 on Saturday
The operator specifies conditions such as 4:00 to 17:00. In response to this designated condition, the data extraction unit 8 extracts data that meets the condition from the knowledge creation base 7. An example of this data is shown in Table 3.

[0018]

[Table 3]

Further, it is also possible to obtain a rough forecast value by simply forecasting the number of customers and sales of this month from this data and use it for preparing the knowledge base. In this case, first, the growth rate calculation unit 8 calculates the latest actual growth rate. In this calculation, as shown in Table 4, for example, the results of the last three months are compared with the results of the same period last year, and the growth rate is obtained from the ratio.

[0020]

[Table 4]

Next, the rough prediction unit 9 multiplies the data of Table 3 by the elongation rate to obtain a rough predicted value. Table 5 Examples of rough prediction values
Shown in.

[0022]

[Table 5]

3. Knowledge Base Creation In the case of fuzzy reasoning, fuzzy rules and membership functions are required as knowledge.

(1) Creation of Fuzzy Rule In this embodiment, a fuzzy rule is separately set for each day of the week / time zone. The time zone can be set appropriately according to the nature and situation of the store. For example 10:
From 02:00 to 12:00, it is a time zone mainly from the opening of the store to the preparation, 1200: 14:00 is lunch time when the number of customers is most concentrated, 14:00 to 17:00 is idle time mainly for soft drinks, 17 The time from 02:00 to 21:00 may be set to be dinner time.

An example of rule creation will be described. The temperature and weather are set in the IF section (preceding section) of the rule, and TH is set.
The number of customers and sales are set in the EN department (consequent department). A fuzzy label with 3 levels of high, medium and low is set for each item. Then, a rule can be created by writing a circle mark in the rule creation table as shown in FIG. For example, if Rule 1 is described, "If the temperature is high and the weather is good, the sales are high and the number of customers is also high".

(2) Creation of membership function The membership function is set for each fuzzy label of each item of the rule. Examples of membership functions are shown in FIGS. Figure 4 shows the membership function for weather,
5 shows a membership function regarding temperature, FIG. 6 shows a membership function regarding the number of customers, and FIG. 7 shows a membership function regarding sales. In creating this membership function,
A trapezoidal type, a triangular type, a normal curve type, or the like can be arbitrarily selected as the function pattern. Then, the membership function is determined by setting parameters such as the inclination and distribution of the selected pattern and the horizontal axis. When setting this parameter, it is possible to easily and appropriately set the function by referring to the above-mentioned rough predicted value and further taking into consideration the expert of the number of customers / sales forecast such as the person in charge of the store.

4. Customer / Sales Inference Function For example, when inferring the customer / sales for the next day, the operator inputs the weather / temperature of the next day from the input / output unit 3 with reference to the weather forecast or the like. The fuzzy inference unit 5 appropriately selects a rule to be used based on the day of the week and the time of day, and performs the fuzzy inference by a method such as the maximum / minimum method. That is, using the input weather and temperature, the degree of establishment of each rule is obtained from the membership function of the rule antecedent part, for example, by the head cut method. Then, the membership functions of the consequent part of each rule are truncated based on the degree of success, and then the centers of gravity of these functions are combined to obtain a conclusion (number of customers / sales).

Further, by comparing the inference result with the actual result, the fuzzy rule and the membership function can be modified so as to be more appropriate, thereby improving the prediction accuracy.

[0029]

As described above, according to the present invention, it is possible to predict the number of customers and sales by computer system by introducing fuzzy reasoning. Therefore, the accuracy of forecasting the number of customers and sales will be improved, and smoother store operations such as appropriate staffing will be possible. Moreover, even if you are not an expert in forecasting the number of customers and sales, you can easily forecast,
The burden on store managers is reduced.

Further, since past performance data is aggregated to obtain a rough predicted value and the predicted value is used as reference data for preparing the knowledge base, fuzzy rules and membership functions can be prepared easily and appropriately. Become.

[Brief description of drawings]

FIG. 1 is a block diagram showing an overall configuration example of a customer number / sales forecasting system according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing an example of past result totalization processing.

FIG. 3 is an explanatory diagram showing a rule creation table.

FIG. 4 is a graph showing a membership function regarding weather.

FIG. 5 is a graph showing a membership function regarding temperature.

FIG. 6 is a graph showing a membership function regarding the number of customers.

FIG. 7 is a graph showing a membership function regarding sales.

[Explanation of symbols]

 1 ... Statistical processing device 2 ... Customer number / sales forecasting system 3 ... Input / output unit 4 ... Knowledge base creation support unit 5 ... Fuzzy reasoning unit 6 ... Results collection unit 7 ... Knowledge creation base 8 ... Data extraction unit 9 ... Growth rate calculation Part 10 ... Schematic prediction unit 11 ... Knowledge base creation unit 12 ... Knowledge base 13 ... Inference unit

Claims (2)

[Claims] 1. A parameter input section for inputting parameters including weather and weekdays on the forecast day, and a membership function for at least weather, weekdays, number of customers, and sales, and number of customers including weather and day of the week in the antecedent section.・ Knowledge base that stores fuzzy rules with sales as consequent part, using the membership function and fuzzy rules,
A fuzzy reasoning customer number / sales forecasting device comprising a fuzzy reasoning unit that deduces the number of customers / sales from the parameters. 2. A device for constructing a knowledge base of the device according to claim 1 based on an input instruction, wherein the past record data of the number of customers and sales in units of day and time by item including weather and holidays. The past result totaling section that accumulates the past actual result, the past actual result accumulating section that accumulates this past actual result data, the latest period data is extracted from the past accumulated data, and the actual result growth rate indicating the fluctuation tendency of the latest recent period data is obtained. The growth rate calculation unit and the rough prediction value calculation unit that extracts the data of the same period as the forecast target time from the past actual data and corrects the data of the same period with the actual growth rate to obtain the rough forecast value of the number of customers and sales And a knowledge base construction apparatus comprising: an output unit that outputs the rough prediction value as reference data.