JP2006277248A - Method, device and program for predicting amount of order received, - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method and device for predicting the amount of orders received which provide high prediction accuracy by considering the inclination of change in the predicted amount of orders received from the past order reception result independent of the experience of a prediction implementing person. <P>SOLUTION: A condition determination part 11 determines whether or not the future change in the amount of orders received shows stable inclination. When it is determined that stable inclination is shown by the condition determination part 11, a prediction operation part 12 calculates the predicted amount of orders received by an arithmetic operation using predicted change showing the change in the amount of orders received for the lapse of a time and an order reception pace expressing the amount of orders received per unit time which are acquired from the past amount of orders received. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an apparatus for predicting a change in the order amount of a product.

  In general, there are various methods for analyzing market trends including product orders.

  For example, in a simple example, the order pace (for example, the number of orders received per day) is calculated based on the number of orders for the latest fixed unit period (for example, weekly or monthly) of the product for which future orders are predicted. The predicted order volume can be calculated by multiplying the order pace by a period (for example, days) to be predicted. In addition, there are various other examples, and there are also advanced calculation methods using statistical methods.

  However, market trends are constantly changing and there is no absolute formula for forecasting orders. For this reason, companies, research institutions, or departments in each industry select an appropriate prediction method based on the field of the product to be handled and the market size. There are various prediction methods ranging from simple to complex, and the method is selected in consideration of the importance of the predicted numerical value.

  For example, when it is not appropriate to spend a high cost or a long time for predicting the order amount, such as at the department level in a company, a simple method based on the order pace as illustrated above is often used. For short-term forecasting, it is possible to forecast with relatively high accuracy even with a simple method based on the order pace.

  However, in the method of calculating from the latest order pace, there is a large error in long-term forecasts such as the first half, second half, and year. However, if the transition of orders received is stable, it is empirically considered that the prediction accuracy of orders received will increase by considering the transition of orders received in the past. In that case, a method such as correcting the predicted order amount in consideration of a rough tendency obtained from past experience is adopted.

  For example, taking into account the transition of orders received over a relatively long period in the past, orders received from the latest order pace are often corrected based on empirical rules. It is also possible to consider the transition of orders received in the past for similar products such as the old model of the product whose order is to be predicted.

  At a relatively small department level, analysts who receive orders are often involved in sales activities and sales promotion activities that directly touch market trends. Has formed a direct market sense. Therefore, it is often possible to predict with high accuracy based on empirical rules even in analysis by simple methods or unstructured methods.

  However, in this way, it is within the discretion of individual experience to correct the predicted order volume in consideration of the transition of the past order volume, and no clear criteria are shown. For this reason, the method for performing such correction has not been available to everyone easily and appropriately.

On the other hand, a method or an apparatus for predicting the order amount by numerical calculation has been proposed (see, for example, Patent Documents 1 and 2). According to these, it is possible to predict the order amount regardless of whether the analyst has experience or not.
JP 2000-399543 A JP 2000-250888 A

  However, the conventional prediction methods as described in Patent Documents 1 and 2 do not take into account the trend of order volume predicted from past orders received, and as such have been done empirically. The correction could not be made. For this reason, there is a case where it is not possible to obtain a predicted order quantity with sufficient accuracy.

  SUMMARY OF THE INVENTION An object of the present invention is to provide a method and apparatus for predicting an order amount that can improve accuracy in consideration of a trend of order amount predicted from past order results, regardless of the experience of a predictor. It is.

In order to achieve the above object, the order receiving prediction apparatus of the present invention includes:
An order quantity forecasting device that predicts future orders received for products,
A condition determination unit that determines whether or not future orders will show a stable trend;
When it is determined by the condition determination unit to show a stable tendency, a predicted transition representing a change in the amount of orders received over time and an order pace representing orders received per unit time obtained from past orders received And a prediction calculation unit that calculates a predicted order amount by the calculation used.

  Therefore, according to the present invention, the condition determination unit determines whether or not the future order volume transition shows a stable trend, and if it is a stable trend, the prediction calculation unit displays the predicted transition as the latest order pace. Therefore, the predicted order quantity can be predicted with high accuracy from the transition tendency predicted from the past order results, regardless of the experience of the forecaster.

  In addition, the condition determination unit may determine whether or not future orders received show a stable tendency based on numerical values of significant variables of the hedonic approach related to the product.

  According to this, since the condition determination unit determines the stability tendency based on the characteristics of the product using the hedonic approach, it is possible to determine an appropriate stability tendency for each product having different characteristics.

  In addition, the condition determination unit may determine that future orders received show a stable tendency when each numerical value of the significant variable is equal to or greater than a predetermined value and balanced.

  According to this, since the condition determination unit determines a product having a small change in product value due to a change in a specific significant variable, in which each of the significant variables is equal to or more than a predetermined value and balanced, a high accuracy is obtained. You can predict the order volume.

  Moreover, the said prediction calculating part is good also as correct | amending the said prediction transition so that it may converge to the said prediction transition from the said order pace.

  According to this, when the condition determination unit determines that the trend is stable, the prediction calculation unit corrects the prediction transition so that the order pace converges to the prediction transition, so the short-term order receipt amount is obtained with high accuracy. It is possible to predict the amount of orders received with high accuracy by the order transition that can be performed and the forecast transition that shows high forecast accuracy over a long period of time when the trend is stable.

  The prediction calculation unit may further correct the predicted transition corrected to converge the order pace based on an actual measurement period used for calculating the order pace.

  According to this, the prediction calculation unit further corrects the predicted transition corrected to converge the order pace according to the actual measurement period used for calculating the order pace, and uses the obtained predicted transition to predict the predicted order amount. Therefore, it is possible to predict the order quantity with high accuracy by the calculation considering the influence given according to the length of the actual measurement period used for calculating the order pace.

  According to the present invention, the condition determination unit determines whether or not the future order volume transition shows a stable trend, and if it is a stable trend, the prediction calculation unit corrects the predicted transition at the latest order pace. Thus, the predicted order amount is calculated, so that it is possible to predict the order amount with high accuracy from the transition tendency predicted from the past order results, regardless of the experience of the predictor.

  Embodiments for carrying out the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a block diagram showing a configuration of an order receipt prediction apparatus according to this embodiment. Referring to FIG. 1, the order receipt prediction device 10 of this embodiment includes a condition determination unit 11 and a prediction calculation unit 12.

  The condition determination unit 11 determines whether or not the transition of the future order amount shows a stable tendency in the prediction of the order amount.

  At that time, the condition determination unit 11 first determines whether or not the trend is stable based on a condition related to sales of a product for which an order amount is to be predicted (hereinafter referred to as “sales condition”). Here, the stable trend means that there is a high possibility that the transition of future orders will match the past data.

  If it is determined from the sales conditions that the trend is stable, the condition determination unit 11 further determines whether or not the trend is stable from the conditions regarding the properties of the product itself (hereinafter referred to as “product conditions”).

  If it is determined that the sales trend and the product condition are both stable, the condition determination unit 11 determines that future order volume changes indicate a stable trend. On the other hand, when it is determined that the trend is not stable under any of the conditions, the condition determination unit 11 determines that the transition of future orders received does not indicate a stable trend.

  The prediction calculation unit 12 performs an order volume prediction calculation using a calculation method according to the determination result of the condition determination unit 11. When the condition determination unit 11 determines that the trend is stable, the prediction calculation unit 12 obtains the predicted order volume by a calculation method using a future predicted transition and the latest order pace. When the condition determination unit 11 determines that the stability tendency is not exhibited, the prediction calculation unit 12 obtains a predicted order volume by a calculation method using the latest order pace.

  FIG. 2 is a flowchart showing the operation of the order receipt forecasting apparatus according to the present embodiment. Referring to FIG. 2, the condition determining unit 11 of the order receipt predicting apparatus 10 first determines whether or not the sales condition tends to be stable (step 101).

  The sales condition is, for example, a determination condition based on the transition of the order received for the product so far. For products that always require a functional improvement, in general, immediately after the start of sales, the order amount shows a rapid rise and then the degree of increase gradually decreases, and after the order amount reaches a peak, it shows a downward trend.

  Although orders for products tend to show an upward trend at the beginning of the market, the products have not yet been deployed in a wide range of markets, and there are few factors that stabilize the transition thereafter. In addition, after the order volume starts to decline, the attractiveness of the product is often lost, but the pace of decline is not stable, and there is a possibility that it will rise again due to the recovery of attractiveness.

  On the other hand, there is a wide range of product development in the market from the start of the decrease in the amount of orders received until the peak of orders received, so there are few fluctuation factors and the possibility of sudden decline is low. , Stable and likely to match past data. There is a computer device as an example of a product showing such a change in the order amount.

  FIG. 3 is a graph showing changes in the order amount. In FIG. 3, as the time t elapses, the order amount f reaches a peak and then gradually decreases. For this reason, the condition determination unit 11 determines that the order tends to be stable when the received order time derivative df / dt ≧ 0 and the time derivative time derivative d (df / dt) / dt ≦ 0.

  If it is determined that there is a stable tendency from the sales conditions, the condition determining unit 11 next determines whether or not a stable tendency is indicated from the product conditions (step 102).

  The product condition is a determination condition based on whether or not each numerical value of the significant variable in the hedonic approach is greater than a predetermined value and balanced. If the significant variable is equal to or greater than a predetermined value and is balanced, it is determined that the trend is stable. The hedonic approach is based on the hypothesis that product value (for example, price) is an integration of the various characteristics of the product. From historical data on many product characteristics and prices, characteristics that are significant in determining the value of the product. This is a technique for quantifying items (significant variables). The numerical value can be used to determine the price of the product. The characteristic items may include sales points that are difficult to quantify such as the presence or absence of functions, in addition to performance such as computer processing capability.

  For example, the condition determination unit 11 converts the significant variable of the target product into a radar chart, and whether each axis value indicated in the radar chart is equal to or greater than a predetermined value and the polygon is close to a regular polygon equal to or greater than a predetermined threshold value. Judge whether or not. It is possible to determine whether or not it is close to a regular polygon by geometric calculation of vector values. The fact that it is close to a regular polygon indicates that the product value is determined by many equal significant variables, that is, the change in the product value due to the change of a specific significant variable is small. For this reason, it is considered that products with a radar chart close to a regular polygon show a stable trend in orders because there is little change in product value with respect to changes in specific significant variables.

  The condition determining unit 11 may use statistical data published by a public institution or the like when determining a significant variable of the hedonic approach and its numerical value in radar charting. In that case, the condition determination part 11 should just memorize | store the statistical data published by the public organization etc. previously, and should just apply the numerical value of the characteristic item of a product to the significant variable shown by the data.

  FIG. 4 is an example of a radar chart showing the values of significant variables of products. Referring to FIG. 4, the radar chart of product (A) is more balanced with significant variables dx1 to dxn than the radar chart of product (B), and is closer to a regular polygon. Therefore, it can be estimated that the product (A) shows a more stable tendency than the product (B).

  Returning to FIG. 2, when it is determined that a stable tendency is indicated from the product conditions, the prediction calculation unit 12 performs an order volume prediction calculation using the prediction transition and the latest order pace (step 103).

  In this order quantity prediction calculation, the prediction calculation unit 12 first obtains a predicted transition of the order quantity over a certain period from past data. The predicted transition represents a change in the amount of orders received over time over a certain period, and is indicated by a numerical sequence on a time series. An appropriate period for obtaining the predicted transition varies depending on the characteristics of the product and the like, but may be about one year for a computer device, for example. Based on the predicted transition obtained here, the predicted transition to be used for the prediction of the order amount is obtained.

  As a specific calculation for obtaining the basic form of the predicted transition, it is possible to multiply the result of the transition in the previous fiscal year by the ratio of this year. In that case, as the ratio, a value obtained from an initially measured value of this year or an expected value may be used.

  Moreover, the prediction calculation part 12 calculates | requires an order pace based on the newest order volume. The past order amount used to determine the order pace is a period shorter than the forecast transition period. The order pace is a numerical value indicating the amount of orders received per unit period, and is different from the predicted transition over time. For example, the order volume per day may be calculated based on the recent weekly or monthly order volume, and used as the order pace.

  Further, the prediction calculation unit 12 compares the order pace with the basic part of the prediction transition corresponding to the time, and obtains the difference. In general, there is a difference between the basic form of forecast transition and the order pace. In the short term, the order volume will be strongly influenced by the pace of orders, but if there is a stable trend, it is likely that the trend will gradually converge to the basic form of the forecast trend.

  Then, the prediction calculation unit 12 corrects the basic form so that the order amount obtained by the order pace converges to the order amount obtained by the basic form of the prediction transition. Specifically, the difference may be exponentially distributed and added to the basic form. Then, the prediction calculation unit 12 calculates a predicted order receipt based on the obtained prediction transition.

For example, regarding M month for which the order volume is to be predicted, assuming that the estimated value of the order volume in the basic form is B (M) and the difference in the index distribution is Δ (M), the predicted order volume f (M) is f (M). M) = B (M) + Δ (M). Note that the exponential distribution of Δ (M) is indicated by an exponential function λe ^−λx (x ≧ 0, λ> 0). Here, x is the number of months elapsed from the month when the actual measurement was performed to the month M.

  Here, the predicted order amount f (M) may be used as the prediction result as it is, but it may be further corrected in consideration of the actual measurement period.

  The longer the actual measurement period, the more the subsequent changes in orders received are affected by the trend of the actual measurement period. For example, when predicting a long-term trend such as a year unit, if the actual measurement period is a relatively long period such as June, the influence of the latest order pace is relatively strong. In that case, when correcting the basic form of the prediction transition, an operation may be used in which the longer the actual measurement period, the stronger the influence of the actual measurement value.

  For example, for M month, the predicted order receipt F (M) is obtained by F (M) = B (M) × ((∫actual measurement value dt + ∫f (t) dt) / ∫B (t) dt). it can. Note that the soot actual measurement value dt is the sum of orders received up to the latest measurable month. ∫f (t) dt is an estimated value obtained from the estimated value f (M) described above for the sum of orders received for the month after the latest month. ∫B (t) dt is an estimated value obtained from the above-mentioned basic form B (M), which is the sum of orders received for the month that can be actually measured and the subsequent months. According to this equation, if the actual measurement period becomes longer, the ratio of the actual measured value dt in (estimated actual value dt + ∫f (t) dt) increases, and the influence of the actual measured value indicates the predicted order receipt F (M). Strongly reflected in

  Returning to FIG. 2, if it is determined in step 101 or step 102 that the trend is not stable, the prediction calculation unit 12 calculates a predicted order volume by multiplying the latest order pace by the period to be predicted. If it is determined that the trend is not stable, subsequent orders cannot be predicted with a high system, but the average predicted order volume calculated from the latest order pace including the possibility of rising and falling is calculated. Is reasonable.

  As described above, according to the present embodiment, the condition determination unit 11 determines whether or not the transition of future orders received shows a stable trend, and if it is a stable trend, the prediction calculation unit 12 Since the forecast order is calculated by correcting the forecast transition at the latest order pace, it is possible to predict the order quantity with high accuracy from the trend trend predicted from the past order record, regardless of the experience of the forecaster. it can.

  In addition, according to the present embodiment, the condition determination unit 11 uses the hedonic approach to determine the stability tendency based on the characteristics of the product. Therefore, the condition determination unit 11 determines the appropriate stability tendency individually for each product having different characteristics. It can be carried out.

  In addition, according to the present embodiment, the condition determination unit 11 determines that the product exhibits a stable tendency when each of the significant variables of the hedonic approach is greater than or equal to a predetermined value and is balanced. Products with little change in product value due to changes in variables can be judged as stable, and orders can be predicted with high accuracy.

  Further, according to the present embodiment, when the condition determination unit 11 determines that the trend is stable, the prediction calculation unit 12 corrects the predicted transition so that the order pace converges to the predicted transition, and the predicted transition is Predicted orders are calculated using this method, so it is possible to predict orders with high accuracy based on the pace of orders that can obtain short-term orders received with high accuracy and the forecast transition that shows high forecast accuracy over the long term when the trend is stable. It is.

  Further, according to the present embodiment, the prediction calculation unit 12 further corrects the predicted transition corrected so as to converge the received order pace according to the actual measurement period used for calculating the received order pace, and the obtained predicted transition is calculated. Therefore, the predicted order amount can be predicted with high accuracy by an operation taking into consideration the influence given according to the length of the actual measurement period used for calculating the order pace.

  In the present embodiment, as an example of the sales condition, the determination condition based on the transition of the order received of the product so far is shown, but the present invention is not limited to this. As another example, the condition determination unit 11 may use a determination condition based on how many stages the sales route leading to the end user is configured.

  In the case of selling through a route through a plurality of secondary and tertiary stages, it is easier for the manufacturer's order volume to show a stable trend than in the case of selling directly to the end user from the manufacturer.

  Secondary and tertiary developments in product distribution channels mean expansion of sales channels and sales areas, indicating an increase in market penetration and production volume of the entire product. Products that are widely deployed in the market tend to be stable with little fluctuation in the short-term order pace. In addition, when end-user trends are reflected in orders received by manufacturers, they are buffered by multiple stages.

  Further, the order quantity prediction device 10 described above can be realized by storing a software program having processing of each unit in a storage device, and reading and executing the software program by the processor.

  Next, specific examples regarding the hedonic approach in the present embodiment will be described.

According to the Bank of Japan Research and Statistics Bureau, “Server Hedonic Regression Formula-Quality Adjustment Method in Corporate Price Index” (June 2004), in 2003, the following variables of the server are considered significant.
(1) Data processing capacity (CPU clock frequency, cache memory capacity, CPU type, standard number of CPUs)
(2) Expandability (maximum installed memory capacity)
(3) Continuous operation (hot plug)
These are applied to a standard server and a fault-tolerant server, and a radar chart of significant variables is created.

  FIG. 5 is a radar chart showing significant variables in two types of servers. The value of each significant variable is normalized with the price conversion value.

  The standard server is a general-purpose server having functions that are generally required. A fault-tolerant server is a server whose reliability with respect to failure is improved by duplicating each part.

  For this reason, with regard to CPUs and memories (expandability), standard servers are lined up so that they can respond to the latest CPU and memory increasing trends required in the market. On the other hand, fault-tolerant servers also have a lineup, but they may not be up-to-date because they require a certain evaluation period and are not as well organized as general-purpose servers. Therefore, for the significant variables (1) and (2), it can be considered that a slightly higher value can be set in a standard server.

  Further, with regard to the hot plug function, the network card portion and the like are duplicated, and the card can be replaced while the system is operating. On the other hand, a fault-tolerant server has most components multiplexed to support hot plugging, and has very high continuous mobility.

  Therefore, for the significant variable (3), a higher numerical value can be set for the fault-tolerant server.

  Referring to FIG. 5, it can be seen that the fault-tolerant server radar chart 21 is closer to a regular polygon than the standard server radar chart 22. Therefore, it can be said that the fault-tolerant server shows a stable order volume transition for a certain period of time and is cheaper than the standard server. For example, if the threshold value for determining similarity to a regular polygon is between a fault-tolerant server and a standard server, it is determined that the fault-tolerant server shows a stable tendency, and the standard server has a stable tendency. It will be determined that it is not shown.

  Next, another specific example regarding the hedonic approach in the present embodiment will be described. The significant variable of the hedonic approach can be used effectively in a period when it can be assumed that the situation regarding the product when the significant variable is extracted is maintained. In the situation where a new significant variable has occurred in the field, the accuracy of the determination made with the past significant variable decreases.

  Therefore, in a field where a new product with a short life cycle and a new function is put on the market in a short time, the period in which the significant variable of the hedonic approach can be used effectively is short. Here, a notebook personal computer is exemplified as a product having a short life cycle in which a new product is put on the market in about six months.

  According to the Bank of Japan Research and Statistics Bureau, “Personal Computer Hedonic Regression Formula-Quality Adjustment Method in the Corporate Price Index” (September 2004), the significant variables in February 2004 are clock frequency, secondary cache capacity, and installed memory. Capacity, HDD capacity, and resolution. However, the significant variables in September 2004 are the clock frequency, secondary cache capacity, installed memory capacity, HDD capacity, screen size, battery duration, weight, wireless LAN installed, and TV tuner function.

  The resolution that was present in February 2004 will disappear in September 2004, and the screen size, battery duration, weight, wireless LAN, and TV tuner function that were not present in February 2004 will be newly significant variables. It has become. Characteristic items peculiar to notebook personal computers become significant, and the latest TV tuner functions are immediately reflected.

  If the significant variable of the hedonic approach changes in about half a year, the forecast period range for orders received should be within that half year. However, a significant variable of the hedonic approach that changes in a short period means that the life cycle of the product itself is short, so it is considered that there are few problems due to the limited range of the forecast period.

It is a block diagram which shows the structure of the order receipt prediction apparatus by this embodiment. It is a flowchart which shows operation | movement of the order volume prediction apparatus of this embodiment. It is a graph which shows transition of orders received. It is an example of the radar chart which shows the value of the significant variable of a product. It is a radar chart which shows the significant variable in two types of servers.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Order amount prediction apparatus 11 Condition determination part 12 Prediction calculation part 21 Radar chart of fault tolerant type server 22 Radar chart of standard server 101-104 steps

Claims (9)

An order quantity forecasting device that predicts future orders received for products,
A condition determination unit that determines whether or not future orders will show a stable trend;
When it is determined by the condition determination unit to show a stable tendency, a predicted transition representing a change in the amount of orders received over time and an order pace representing orders received per unit time obtained from past orders received An order quantity predicting apparatus comprising: a prediction calculation unit that calculates a predicted order quantity by a calculation used.   The order determination apparatus according to claim 1, wherein the condition determination unit determines whether or not future orders show a stable tendency based on each numerical value of a significant variable of the hedonic approach related to the product.   The order determination unit according to claim 2, wherein the condition determination unit determines that future orders received show a stable tendency when the numerical values of the significant variables are equal to or greater than a predetermined value and are balanced.   The order amount prediction apparatus according to claim 1, wherein the prediction calculation unit corrects the prediction transition so that the prediction transition converges from the order pace.   The order amount prediction device according to claim 4, wherein the prediction calculation unit further corrects the prediction transition corrected to converge the order pace based on an actual measurement period used for calculating the order pace. A method for predicting orders received for predicting future orders received for products,
Determining whether or not future orders will show a stable trend;
Predicted by calculation using forecast transitions representing changes in orders received over time and order paces representing orders received per unit time, obtained from past orders received when it is determined to show a stable trend An order quantity forecasting method comprising the step of calculating an order quantity.   The order receipt prediction method according to claim 6, wherein it is determined whether or not future order receipts show a stable tendency based on each numerical value of a significant variable of the hedonic approach for the product. An order quantity forecasting program for causing a computer to predict the future order quantity of a product,
A procedure to determine whether future orders will show a stable trend,
Predicted by calculation using forecast transitions representing changes in orders received over time and order paces representing orders received per unit time, obtained from past orders received when it is determined to show a stable trend A program for causing a computer to execute a procedure for calculating an order amount.   The program according to claim 8, wherein it is determined whether or not future orders received show a stable tendency based on each numerical value of a significant variable of the hedonic approach for the product.

Images