JPH10228463A - Demand prediction model evaluating method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a demand prediction model evaluation method nor speeding up and facilitating evaluation work verifying availability to a real job by means of the application of a demand prediction model, and suppressing a wasteful prediction model reconstruction work that is to secure precision more than that requested in a job system. SOLUTION: The evaluation method is constituted of processing procedures for taking in a demand result value in the evaluation period of the demand prediction model and a demand prediction value, calculating the deviation value of the demand result value and the demand prediction value at every supply cycle of a product to a market, which is inputted from an input device, calculating supply excess cost and supply lack cost in accordance with the deviation value and outputting supply excess cost and supply lack cost to an output device.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to demand forecasting performed as a preceding stage of sales planning, production planning, etc., and is particularly suitable for demand forecasting when evaluating the usefulness of an established demand forecasting model for actual work. Model evaluation method.

[0002]

2. Description of the Related Art In demand forecasting, a forecasting model in which demand characteristics are formulated by various statistical and forecasting methods is used in order to improve forecasting accuracy. In constructing this prediction model, an approach is often taken to narrow down an appropriate prediction model to be used in the business from a plurality of prediction models that can be considered as alternatives according to a reference evaluation index. At that time, AIC (An Information Criterion) is one of the representative indices adopted as evaluation indices. Literature
(Hiromatsu, Namika: Economic Time Series Analysis, Asakura Shoten, 1990.
According to 11, pp58), the AIC is an index for comparing the poorness of a model having a parameter determined by the maximum likelihood method, and the larger the AIC is for the same data, the worse the model is. Therefore, from the standpoint of evaluating the prediction model, attention should be paid to the magnitude of AIC.

[0003]

In the above prior art, the quality of a prediction model is evaluated only by an evaluation index from a statistical / prediction theoretical viewpoint. Therefore, in order to clearly explain the effects of applying a prediction model to people other than experts in statistics and prediction, evaluation results that take into account the effect of improving the prediction accuracy by applying a prediction model on business systems will be added. Was necessary. Furthermore,
In the above prior art, since only the AIC is a criterion for determining whether or not the prediction model should be further refined, an excessive prediction model rebuilding operation is likely to occur in order to secure more accuracy than required by the business system. There was a problem.

SUMMARY OF THE INVENTION In view of the above problems, it is an object of the present invention to speed up and facilitate an evaluation work for verifying the usefulness of a demand forecasting model for an actual business, and to provide a business system which is more than required. It is an object of the present invention to provide a demand forecasting model evaluation method that suppresses useless forecasting model reconstruction work for ensuring accuracy.

[0005]

The above object is achieved by the following means (1) to (7).

(1) An input device, an output device, a storage device including a storage area for information relating to the actual demand value of a product and a storage area for storing information relating to the predicted demand value of the product calculated by a demand prediction model, and the storage device In a demand forecasting model evaluation method using a computer having an arithmetic unit that calculates an evaluation value of a demand forecasting model in an evaluation period input from the input device based on information in the device, an actual demand value in the evaluation period and Incorporating demand forecast values, calculating a divergence value between the actual demand value and the demand forecast value for each supply cycle of a product to the market input from the input device, and determining an excess supply cost and supply according to the divergence value. A shortage cost is calculated and the excess supply cost and the insufficient supply cost are output to the output device.

(2) If the value obtained by subtracting the actual demand value from the demand forecast value in the supply cycle is positive, it is determined that the supply of the product to the market is excessive, and then the cost associated with the excess supply is calculated. It is characterized by the following.

(3) If the value obtained by subtracting the actual demand value from the demand forecast value in the supply cycle is smaller than the limit value that can be stocked in the market, it is classified into inventory holding cost, and if larger, it is classified into disposal cost. Calculate the costs associated with oversupply.

(4) If the value obtained by subtracting the actual demand value from the predicted demand value in the supply cycle is negative, it is determined that the supply amount of the product to the market is insufficient, and then the cost associated with the supply shortage is calculated. .

(5) If the value obtained by subtracting the actual demand value from the demand forecast value in the supply cycle is smaller than the limit value for urgent procurement in the market, it is classified into emergency procurement cost, and if larger, it is classified into opportunity loss cost. And calculate the cost associated with the supply shortage.

(6) The difference between the actual demand value and the demand forecast value for each supply cycle is regarded as the theoretical stock, and the time series transition of the theoretical stock from the start to the end of the demand forecast model evaluation period The figure is output to an output device.

(7) In the time series transition diagram of the theoretical inventory, the excess inventory and the short inventory are identified by changing the output form according to the sign of the difference between the actual demand value and the predicted demand value.

[0013]

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

FIG. 1 is an apparatus configuration diagram of the present embodiment, which comprises an input device 1, an output device 2, a storage device 3, an arithmetic device 4, and a control device 5 for controlling them. In the storage device 3, as shown in FIG. 2, a sales volume actual data indicating a predicted target product name storage column 201 and a sales volume performance of the prediction target product (in this embodiment, the sales volume is regarded as a demand volume). Column 20
2, a forecast target product name storage column 301 as shown in FIG. 3, a forecast value data column 302 indicating a value obtained by predicting the sales volume of the forecast target product by the demand forecast model. Forecast information table 30 having
0, an evaluation period table 400 indicating the evaluation period of the demand forecast model input from the input device 1 is stored as shown in FIG. The arithmetic unit 4 includes an arithmetic function 6 for performing four arithmetic operations and various function operations, and a demand prediction model evaluation function 7 for evaluating the effect of applying the demand prediction model based on the present invention. Hereinafter, the processing operation and the information transmission operation of the demand forecasting model evaluation function 7, which is the center of the present invention, will be described with reference to the flowchart of FIG. It should be noted that the forecast value data used in the present embodiment has already been calculated for each product by the demand forecast model, and the demand forecast model (hereinafter abbreviated as “forecast model”) used for the calculation has been optimized for the model structure and structural parameters. Assume that it is.

<Step 501> Import of sales performance / sales prediction data Sales performance information table 20 stored in storage device 3
0 and the sales forecast information table 300, the sales performance data and the sales forecast data of the product to be predicted are fetched, and stored in the evaluation target data table 600 in the format shown in FIG.

<Step 502> Incorporation of Conditions for Evaluating Predictive Model As conditions necessary for evaluating the predictive model, first, as shown in FIG. 7, a market (hereinafter, the market is treated as a stock storage warehouse in this embodiment) is shown in FIG. In response to this, a supply cycle indicating the intervals at which the products are supplied is taken from the input device 1, and a supply cycle table 800 of the type shown in FIG.
To be stored. Further, a cost calculation parameter for calculating how much a loss cost is generated when a warehouse is restocked in accordance with a predicted value by a prediction model and the supply cycle described above is fetched from the input device 1 and has a format shown in FIG. Is stored in the cost calculation parameter table 900. Here, the cost calculation parameter table 900 includes, as information 901 for excess supply cost calculation, a stock holding limit value indicating how much stock can be held in the warehouse when the supply amount exceeds the demand amount. 902, a stock holding unit cost 903 indicating a cost required for stock holding per unit amount when the deviation amount is larger than the demand amount when the supply amount exceeds the demand amount 902, and a deviation when the supply amount exceeds the demand amount. A disposal unit cost 904 indicating the cost required for product disposal per unit amount generated when the amount exceeds the stock holding limit value 902 is stored, and the supply amount is lower than the demand amount as the supply shortage cost calculation information 905. The emergency procurement limit value 906 indicates to what extent emergency procurement of the product is possible, and the absolute value of the deviation amount when the supply amount falls below the demand amount is urgently adjusted. Emergency procurement unit cost 907 indicating the cost required for urgent procurement per unit amount below the limit value 906, a unit generated when the absolute value of the deviation amount exceeds the emergency procurement limit value 906 when the supply amount falls below the demand amount Opportunity loss unit cost 90 indicating opportunity loss per amount
8 is stored.

<Step 503> Calculation of deviation between supply amount and demand amount The deviation amount between supply amount and demand amount in the supply cycle is calculated. Specifically, as shown in FIG.
The sales performance data and the sales forecast data in the evaluation target data table 600 are totaled for each supply cycle in the supply cycle table 800 for the period from the evaluation start time to the evaluation end time of the table 400. After obtaining the cumulative sales value J (t) and the cumulative sales forecast value Y (t) at the supply time t, the deviation amount R (t) between the supply amount and the demand amount is calculated by the following equation. This is stored in the deviation amount table 1100 shown in FIG.

R (t) = Y (t) -J (t) <Step 504> Calculation of loss cost Assuming a case where inventory is replenished to the warehouse in accordance with the predicted value by the prediction model and the set supply cycle, the supply cycle is assumed. The loss caused by the divergence between the supply amount and the demand amount is calculated as the cost. The specific processing procedure will be described with reference to the flowchart of FIG.

First, the difference R (t) between the supply amount and the demand amount at the supply time t is fetched from the difference amount table 1100 (step 1201), and the type of the loss cost is determined based on the magnitude of the difference R (t). A determination is made (step 1202). If the deviation R (t) is positive and exceeds the stock holding limit value 902, the disposal cost S (t) is calculated by the following equation (step 12).
03).

S (t) = K1 * C1 + (R (t) −K1) * C2 where K1: inventory holding limit value 902 C1: inventory holding unit cost 903 C2: disposal unit cost 904 Deviation amount R (t) Is positive and equal to or less than the stock holding limit value 902, the stock holding cost Z (t) is calculated by the following equation (step 1204).

Z (t) = R (t) * C1 The deviation R (t) is negative and its absolute value is the emergency procurement limit value 9
If it is not more than 06, the emergency procurement cost W (t) is calculated by the following equation (step 1205).

W (t) = | R (t) | * C3 where C3: emergency procurement unit cost 907 The deviation R (t) is negative and its absolute value is the emergency procurement limit value 906.
If the value exceeds the value, the opportunity loss cost V (t) is calculated by the following equation (step 1206).

V (t) = K2 * C3 + (| R (t) | -K2) * C4 where K2: emergency procurement limit value 906 C4: opportunity loss unit cost 908 As described above, the deviation amount R (t) After calculating the loss cost at the supply time t according to the magnitude of the
The cost obtained by adding the cumulative loss cost up to -1) is stored in the cumulative loss cost table 1300 of the format shown in FIG. 13 as the cumulative loss cost up to the supply time t (step 1207). Finally, it is determined whether or not the calculation of the loss cost has been completed at all supply points in the period from the evaluation start point to the evaluation end point stored in the evaluation period table 400 (step 1208). If so, the supply point is updated and the process returns to step 1201, and if completed, this step ends.

<Step 505> Output of prediction model evaluation result The contents of the accumulated loss cost table 1300 are output to the output device 2. Specifically, as shown in FIG. 14, the loss cost when the evaluation target prediction model is applied, and the reference prediction model to be compared (here, it is assumed that the demand at each point in the evaluation period is constant). Using a model that calculates the predicted value based on
, A cost reduction information output area 1401 showing how much the evaluation target prediction model contributes to cost reduction compared to the reference prediction model.
The screen is constructed around 2. The cost reduction amount CG and the cost reduction rate CR in the cost reduction information output area 1402 are calculated by the following equations.

CG = M−NCR = ((M−N) / M) * 100 where N: loss cost when the evaluation target prediction model is applied M: loss cost when the reference prediction model is applied When the stock transition diagram button 1403 is clicked in the same figure, the difference between the supply amount and the demand amount for each supply cycle is regarded as theoretical stock, and the target from the start time to the end time of the prediction model evaluation period is considered. Then, a time series transition diagram of the theoretical stock is output to the output device 2. FIG. 15 shows an example of the output. In the figure, different output forms are used for the case where the theoretical stock is positive and the case where the theoretical stock is negative, so that it is easy to distinguish between the excess stock and the short stock. As the identification method, there is a method of changing the type of hatching as in the present embodiment, but other methods of changing the type of color, fill, and the like are also conceivable.

<Step 506> Judgment of Completion of Prediction Model Evaluation Processing If the prediction model evaluation result output in step 505 is satisfactory, the present step is terminated. Change the conditions and evaluate the prediction model again.

Through the above-described processing steps, it is possible to quantitatively calculate the loss cost relating to the excess supply and the shortage of supply that may occur by applying the prediction model, and to present the result to the predictive model evaluator in an easy-to-understand manner. As a result, the prediction model evaluation work can be speeded up and facilitated, and it can be expected that useless prediction model rebuilding work for securing accuracy higher than required by the business system is suppressed.

[0028]

According to the present invention, it is possible to speed up and simplify the operation of verifying the usefulness of the actual business by applying the demand forecasting model, and to secure the accuracy higher than required by the business system. Useless prediction model rebuilding work can be suppressed. As a further side effect, the present invention makes it easy to understand the usefulness of the demand forecasting model by applying it to actual work. It can be expected to increase for end users.

[Brief description of the drawings]

FIG. 1 is an apparatus configuration diagram of an embodiment.

FIG. 2 is a configuration diagram of a sales performance information table relating to a product to be predicted.

FIG. 3 is a configuration diagram of a sales prediction information table relating to a product to be predicted.

FIG. 4 is a configuration diagram of a demand prediction model evaluation period table.

FIG. 5 is a diagram for explaining a processing operation and an information transmission operation of the demand prediction model evaluation method.

FIG. 6 is a configuration diagram of an evaluation target data table.

FIG. 7 is a diagram for explaining a supply cycle.

FIG. 8 is a configuration diagram of a supply cycle table.

FIG. 9 is a configuration diagram of a cost calculation parameter table.

FIG. 10 is a diagram for explaining a deviation amount between a supply amount and a demand amount.

FIG. 11 is a configuration diagram of a deviation amount table.

FIG. 12 is a flowchart showing a procedure for calculating a loss cost.

FIG. 13 is a configuration diagram of a cumulative loss cost table.

FIG. 14 is a diagram illustrating an example of an output screen of a prediction model evaluation result.

FIG. 15 is a diagram illustrating an example of an output screen of a time-series transition diagram of theoretical stock.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Input device, 2 ... Output device, 3 ... Storage device, 4 ...
Arithmetic unit, 7: predictive model evaluation function, 502: capture of predictive model evaluation conditions, 504: calculation of loss cost, 505: output of predictive model evaluation results

Claims (7)

[Claims] 1. An input device, an output device, a storage device including a storage area for information relating to a demand actual value of a product and a storage area for storing information relating to a demand prediction value of the product calculated by a demand prediction model, and the storage device A demand prediction model evaluation method using a computer having an arithmetic unit that calculates an evaluation value of a demand prediction model in an evaluation period input from the input device based on information in the demand period. A predicted value is fetched, a deviation value between the actual demand value and the predicted demand value is calculated for each supply cycle of the product to the market input from the input device, and an excess supply cost and an insufficient supply are determined according to the deviation value. A demand prediction model evaluation method, comprising calculating a cost and outputting the excess supply cost and the insufficient supply cost to the output device. 2. The demand prediction model evaluation method according to claim 1, wherein when the value obtained by subtracting the actual demand value from the demand forecast value in the supply cycle is positive, it is determined that the supply amount of the product to the market is excessive. And then calculating the cost associated with excess supply. 3. The demand forecasting model evaluation method according to claim 2, wherein if a value obtained by subtracting the actual demand value from the demand forecast value in the supply cycle is smaller than a limit value that can be stocked in the market, an inventory holding cost; A demand forecasting model evaluation method characterized in that when it is large, it is classified into disposal costs and the costs associated with excess supply are calculated. 4. A demand forecasting model evaluation method according to claim 1, wherein if the value obtained by subtracting the actual demand value from the demand forecast value in the supply cycle is negative, it is determined that the supply amount of the product to the market is insufficient. And calculating a cost associated with a shortage of supply. 5. The demand forecasting model evaluation method according to claim 4, wherein when a value obtained by subtracting the actual demand value from the demand forecast value in the supply cycle is smaller than a limit value for urgent procurement in the market, the emergency procurement cost. A demand prediction model evaluation method characterized by classifying the costs into opportunity loss costs when the costs are large, and calculating costs associated with supply shortages. 6. The demand forecasting model evaluation method according to claim 1, wherein a deviation value between the actual demand value and the demand forecasting value for each supply cycle is regarded as theoretical stock, and the demand forecasting model evaluation period is started and ended. A demand forecasting model evaluation method characterized by outputting a time-series transition diagram of theoretical stock to an output device for up to a time point. 7. The demand forecasting model evaluation method according to claim 6, wherein the time series transition diagram of the theoretical inventory is determined by changing the output form according to the sign of the difference between the actual demand value and the predicted demand value to thereby determine the excess inventory. A demand forecast model evaluation method characterized by identifying shortage stock.