KR101502539B1 - Method and apparatus for obtaining optimal location and capacity of expressway service area - Google Patents

Abstract

The method for calculating the optimum size of a highway rest area according to an embodiment of the present invention is characterized in that each case includes at least a case database represented by a plurality of independent variables including a passing traffic volume, Providing a test case for which variable test values are given, estimating a restaurace utilization estimate based on the results retrieved from the case database for a given independent variable test values, and estimating a given independent variable test values and resting place utilization estimates Estimating a sales forecast value per vehicle based on the result retrieved from the case database. Further, the step of determining the scale of the convenience store facility capable of achieving the target service rate from the rest area utilization rate prediction value may be further included.

Description

TECHNICAL FIELD [0001] The present invention relates to a method and an apparatus for calculating an optimum position and a size of a highway rest area,

BACKGROUND OF THE INVENTION 1. Field of the Invention [0002] The present invention relates to a highway rest area operation, and more particularly, to a position and scale optimization technique of a highway rest area.

According to the road design manual stipulating the design standards of expressways, the expressway rest area is "a facility that can relieve the fatigue and tension of continuous high-speed driving, resolve driver's physiological needs, and provide opportunities for refueling and maintenance of automobiles ".

The rest area should include parking lots, green spaces, restrooms, restaurants, kiosks, gas stations, garages, auxiliary facilities, squares, corridors, lanes, etc. However, gas stations, kiosks and restaurants are not required.

Regarding the installation of a highway rest facility, the road design manual recommends that "the distance between each facility should be set so as not to be too close or too far away" and that various facilities such as resting places, Standard 15 km, and a maximum of 25 km away from each other, and 50 km between the service stations and a maximum of 100 km, and 50 km between the service stations and 60 km away from each other.

The reason and reasonableness of the spacing of these resting places are unclear, but the fact that many long-distance resting places are needed for long-distance roads remains unchanged. In fact, in the case of the Republic of Korea, the highway rest area was opened as a total of 170 as of June 2011 after four rest areas were opened in 1970. However, 28 rest areas were closed due to deficit management and 142 as of June 2011 are in operation. According to reports, about 25% of rest areas are recording losses.

The reason why many highway rest areas are operated or closed down than the general observations is that the investment plan may be due to lack of interest, but the current regulations may be inadequate in three aspects as follows.

First, it is a matter of estimating the size of the rest area. The rest area is basically a facility for relieving the driver's rest and physiological needs, so a large parking space is needed. The current regulation calculates the number of parking spaces according to the amount of traffic passing through and the number of parking spaces according to the number of parking spaces, the facilities (restrooms, green spaces, etc.), sales facilities (gas stations, kiosks, restaurants, etc.) and operating facilities (management offices, sewage treatment plants, Etc.) are collectively calculated. For example, referring to Table 902.7 of the road design manual, when the number of parking lots is 251 or more, the standard size of rest area is defined as 350 sq. M., 950 sq. M., 255 sq. M., And 550 sq.

Second, pass traffic volume is not applied in the calculation process of parking lots and is applied excessively. In the current regulation, design criteria defines the top 10% of the 365 days of transit traffic recorded at 10 years after opening, that is, the 35th pass traffic volume. This is defined as the capacity to secure service for the resting place on the use day of 90% per year even if the service capacity is exceeded at the use of 10% per year. These regulations are designed to ensure that services are smoothly used on the day of use of the top 10%, even though the traffic volume fluctuates greatly during a year and the utilization rate is low on most use days. For example, Forces to maintain facilities that may be possible.

Third, the conversion factor applied in the calculation of the number of parking spaces, the usage rate of rest area, the turnover rate, and the congestion rate can be calculated by indiscriminately introducing the overseas standard or using the limited survey results, or applying the fixed value And does not adequately reflect reality. For example, it is expected that the facility utilization rate will be low in a place close to the departure city or the destination city, but the utilization rate and the congestion rate indicated in the regulations are applied equally to any location. As a result, the number of parking lots is much larger than the actual demand and the size of the facility is also calculated to be larger than the actual demand. In this case, in order to obtain a business license, it is necessary to invest a lot of sales in the sales facility, which leads to a deficit.

Even though rest areas require more than KRW 10 billion to invest in installation, there is a limit to the improvement of profits even if deficits are brought about due to limitations on intellectual property. It may be unfair to pass all of these problems to private operators who have the right to operate rest areas.

In the first place, the road design manual suggests the conditions of natural environmental conditions, securing of paper and facilities, convenience of construction and maintenance, conditions of road and traffic, and disregarding the location condition of economic analysis.

Statistically, according to the results of analyzing the use of highway rest areas in Korea by the inventor as of the filing date, the correlation coefficient between the passing traffic volume and resting area construction area is about 0.38, the correlation coefficient between the number of resting place vehicles and resting area construction area Is 0.43, which is very low correlation. It is statistically proven that the method of estimating the building size of rest areas according to the current regulations is very unreasonable.

Despite the recognition of these problems, existing researches have been carried out on the calculation of the conversion factors such as the utilization rate of resting place, turnover rate, congestion rate, and numerical calculation of the number of parking spaces, Because it is limited to research, various characteristics of the rest area are not reflected, so there is a limit to predict the appropriate scale.

In addition, some studies have been conducted on investment decision making through forecasting of sales, but there is a problem in the reliability of the model, such as using a fixed conversion coefficient, or limited to a linear model such as regression analysis.

Therefore, in order to improve the irrationality of the current highway rest area estimating model and to prevent the opening of a rest area, it is necessary to accurately predict the appropriate scale according to the planned location of the rest area.

Furthermore, in order to increase profitability and efficiency even at the same scale, a model is required to accurately calculate the detailed building area from the utilization rate of the facility, instead of collectively calculating the parking lot size as in the current regulations.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and it is an object of the present invention to provide a method and an apparatus for calculating an optimum position and a size of a highway rest area to be newly constructed or enlarged based on operational cases collected so far.

According to an aspect of the present invention, there is provided a method for calculating an optimum size of a highway rest area,

Providing a test case in which each case is represented by a plurality of independent variables including at least a passing traffic volume, a case database represented as a dependent variable including resting place utilization and sales per vehicle, and a test case given independent variable test values;

Estimating a rest area utilization rate prediction value based on a result retrieved from the case database with respect to the given independent variable test values; And

Estimating a sales forecast value per vehicle based on a result retrieved from the case database with respect to the given independent variable test values and the rest area utilization predicted value.

In one embodiment, estimating the rest area utilization rate prediction value includes:

Dividing the resting place utilization rate values of the cases in the case database into at least two clusters according to the passing traffic criterion;

Searching for similar cases with respect to the given independent variable test values in a cluster to which the given passing traffic belongs; And

And estimating a rest area utilization rate prediction value by referring to rest area utilization values of the retrieved similar cases.

In one embodiment, estimating the per-vehicle sales forecast value comprises:

Dividing the sales values per vehicle of the cases in the case database into at least two clusters according to a highway number reference;

Retrieving similar cases with respect to the independent variable test values and the rest area utilization predictive value given in the community to which the highway to which the rest area is to be installed belongs; And

And estimating a sales forecast value per vehicle by referring to sales values per vehicle of the retrieved similar cases.

In one embodiment, the retrieval of the case database comprises:

Determining a filtering range based on respective dependent variable predictions and error rates based on at least two quantitative prediction methodologies;

Optimizing at least parameters for case based reasoning using a Gene Algorithm; And

And outputting the search cases whose dependent variable values are within the filtering range among the search cases obtained from the case-based reasoning based on the optimized parameters as a similar case.

In one embodiment, the at least two quantitative prediction methodologies may include Multiple Regression Analysis (MRA), ANN (Artificial Neural Network).

In one embodiment, the filtering range is

Range between a maximum of the lower bounds of the prediction ranges obtained from the at least two prediction methodologies and a minimum of the upper bounds of the prediction ranges obtained from the at least two prediction methodologies.

In one embodiment, the filtering range is

An intersection range between a maximum of the lower bounds of the prediction ranges obtained from the at least two prediction methodologies and a minimum of the upper bounds of the prediction ranges obtained from the at least two prediction methodologies by an allowance tolerance range Lt; / RTI >

In one embodiment, the parameters for the at least case-

As the parameters for case-based reasoning, MCAS (Range Criteria for Scoring Attribute Similarity), RAWn (Range of Case Attribute Weight), RCS (range of case selection) and parameters for determining the filtering range, Tolerance range of cross range between the predicted values of MRA and ANN models.

In one embodiment, the partitioning of the clusters may be performed according to a cluster analysis methodology including a decision tree methodology.

In one embodiment, the optimal size calculating method of the highway rest area includes:

And determining a scale of a rest area facility capable of achieving a target service rate from the rest area utilization rate predicted value.

In one embodiment, the optimal size calculating method of the highway rest area includes:

If the sales per unit area is selected based on the probability distribution function of the sales per unit area derived from the cases stored in the case database, the sales facility size of the rest area where the sales value pertaining to the rest area sales can be achieved with the sales per unit area And a step of deciding whether or not to perform a search.

In one embodiment, the method for calculating the optimal size of the highway rest area includes

And estimating the area of the convenience store including the parking lot of the highway rest area based on the predicted value of the rest area utilization rate.

According to another aspect of the present invention, there is provided a method for calculating an optimal input resource of a test case project,

Providing a test case in which each case is presented with at least independent variables and first and second target variables, and an independent variable test value;

Estimating a first dependent variable predictive value based on a result retrieved from the case database with respect to the given independent variable test values; And

Estimating a second dependent variable predicted value based on a result retrieved from the case database with respect to the given independent variable test values and the first dependent variable predicted value.

In one embodiment, estimating the first dependent variable predictive value comprises:

Dividing the first dependent variable values of the cases in the case database into at least two clusters according to a first split criterion independent variable;

Searching for similar cases with respect to the given independent variable test values in the cluster to which the test value of the first split criterion independent variable belongs; And

And estimating a first dependent variable predicted value by referring to the first dependent variable values of the retrieved similar cases.

In one embodiment,

Wherein estimating the second dependent variable predicted value comprises:

Dividing the second dependent variable values of the cases in the case database into at least two clusters according to a second split criterion independent variable;

Searching similar cases for the given independent variable test values and the first dependent variable predicted value in a cluster to which the test value of the given second split criterion independent variable belongs; And

And estimating a second dependent variable predicted value by referring to the second dependent variable values of the retrieved similar cases.

In one embodiment, the retrieval of the case database comprises:

Determining a filtering range based on respective dependent variable predictions and error rates based on at least two quantitative prediction methodologies;

Optimizing parameters for case based reasoning using a Gene Algorithm; And

And outputting the search cases whose dependent variable values are within the filtering range among the search cases obtained from the case-based reasoning based on the optimized parameters as a similar case.

In one embodiment, the at least two quantitative prediction methodologies may include Multiple Regression Analysis (MRA), ANN (Artificial Neural Network).

In one embodiment, the filtering range is

Range between a maximum of the lower bounds of the prediction ranges obtained from the at least two prediction methodologies and a minimum of the upper bounds of the prediction ranges obtained from the at least two prediction methodologies.

In one embodiment, the filtering range is

An intersection range between a maximum of the lower limit values of the prediction ranges obtained from the at least two prediction methodologies and a minimum of the upper limit values of the prediction ranges obtained from the at least two prediction methodologies may be an enlarged intersection range enlarged by an allowable marginal rate .

In one embodiment, the partitioning of the clusters may be performed according to a cluster analysis methodology including a decision tree methodology.

In one embodiment, an optimal input resource calculation method of the test case project includes:

Deriving a probability distribution curve of performance based on the input resource from the cases stored in the case database when the second dependent variable value is related to the performance, Determining the amount of input resources to be input to achieve the second dependent variable predicted value in the project.

According to yet another aspect of the invention, an optimal input resource calculating device of a test case project comprises:

A case database in which each case is represented by at least independent variables and first and second dependent variables;

A first predictor for estimating a first dependent variable predictive value based on a result retrieved from the case database with respect to the given independent variable test values, when a test case given with independent variable test values is provided; And

And a second predictor for estimating a second dependent variable predicted value based on a result retrieved from the case database with respect to the given independent variable test values and the first dependent variable predicted value.

According to one embodiment, the optimal input resource calculating device of the test case project includes:

And dividing the dependent variable values of the cases in the case database into at least two clusters according to the split criterion independent variable and providing the divided clusters to the first predictor or the second predictor.

In one embodiment, each of the first predictor and the second predictor includes:

A quantitative prediction analyzer for computing dependent variable predictive values, error rates and prediction ranges with respect to the cases of the case database based on at least two quantitative prediction methodologies;

A case based reasoning model for searching cases similar to test cases using case based reasoning among the cases of the case database;

A parameter optimizer for optimizing parameters for the case based reasoning model using a genetic algorithm; And

And a filtering unit for filtering the cases retrieved from the case-based reasoning model with a filtering range determined based on the dependent variable predictive values, error rates, and prediction ranges, and outputting a similar case.

In one embodiment, the optimal input resource calculating device of the test case project includes:

A resource-performance probability distribution analyzing unit for analyzing an empirical probability distribution on the performance per input resource based on the cases of the case database; And

And an optimal input resource determiner for analyzing resources that a decision maker has to input to achieve the second dependent variable predicted value with a specific target performance based on the empirical probability distribution.

According to the method and apparatus for calculating the optimum position and size of a highway rest area according to the present invention, the utilization ratio of resting place is predicted from the passing traffic amount, and the actual parking demand can be predicted because the number of parking spaces is calculated from the resting place utilization rate.

According to the method and apparatus for calculating the optimum location and size of a highway rest area according to the present invention, since the sales amount is calculated from the resting place utilization rate and the building area is calculated based on the sales amount, the area of the sales facility that can actually be operated effectively, You can predict sales realistically.

According to the method and apparatus for calculating the optimum location and size of a highway rest area according to the present invention, it is possible to improve the low accuracy and high deviation that are problematic in the conventional case-based reasoning, and to calculate the more accurate reasoning result with fast operation, high accuracy, .

According to the method and apparatus for calculating the optimum location and size of a highway rest area according to the present invention, it is possible to present the optimal scale and the distribution of sales probability in various cases to decision makers based on fast calculation speed and high accuracy. Decision-makers can choose the location and size of the most desirable resting place, based on criteria such as the highest turnover in the regulation or the highest return on investment.

FIG. 1 is a conceptual diagram illustrating a two-step optimization procedure for calculating a rest area utilization rate and a sales amount in a method for calculating an optimal size of a highway rest area according to an embodiment of the present invention.
FIG. 2 is a flowchart illustrating concrete steps of an optimal input resource calculating method of a test case project, which is an example of calculating an optimal scale of a highway rest area according to an embodiment of the present invention.
FIG. 3 is a flowchart illustrating concrete steps of a hybrid case-based reasoning methodology applied in an optimal input resource calculation method of a test case project, which is an example of calculating an optimum size of a highway rest area according to an embodiment of the present invention.
FIG. 4 is a conceptual diagram illustrating a chromosome of a genetic algorithm (GA) in a method for calculating an optimum input resource of a test case project, which is an example of calculating an optimum size of a highway rest area according to an embodiment of the present invention.
FIG. 5 is a conceptual diagram illustrating a first cluster partitioning method for dividing a cluster of resting place utilization rates according to the amount of passing traffic in a method for calculating an optimum input of a test case project, which is an example of calculating an optimal size of a highway rest area according to an embodiment of the present invention to be.
FIG. 6 illustrates a second cluster partitioning method for dividing a cluster of sales per vehicle according to a highway number in an optimal input resource calculating method of a test case project, which is an example of calculating an optimal size of a highway rest area according to an embodiment of the present invention It is a conceptual diagram.
FIG. 7 is a graph illustrating a probability distribution function for achieving sales per unit area in an optimal input resource calculating method of a test case project, which is an example of calculating an optimum size of a highway rest area according to an embodiment of the present invention.
FIG. 8 is a block diagram illustrating an optimal input resource calculating apparatus of a test case project illustrating an optimum scale calculation of a highway rest area according to an embodiment of the present invention.

For the embodiments of the invention disclosed herein, specific structural and functional descriptions are set forth for the purpose of describing an embodiment of the invention only, and it is to be understood that the embodiments of the invention may be practiced in various forms, The present invention should not be construed as limited to the embodiments described in Figs.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. The same reference numerals are used for the same constituent elements in the drawings and redundant explanations for the same constituent elements are omitted.

FIG. 1 is a conceptual diagram illustrating a two-step optimization procedure for calculating a rest area utilization rate and a sales amount in a method for calculating an optimal size of a highway rest area according to an embodiment of the present invention.

Referring to FIG. 1, in order to calculate the optimal size of the rest area of the expressway, a predicted value of resting place utilization rate is estimated based on attributes such as the passing traffic amount in the first step, and a predicted sales value per vehicle is estimated based on the estimated estimated resting place utilization rate do. The predicted sales value of the resting place is calculated from the predicted value of the sales per vehicle and the utilization rate of the resting place (that is, the number of the resting place vehicles in the passing traffic volume).

The rest area construction area required to achieve the resting place sales forecast value can be decided by the decision maker from the probability distribution curve of the sales per unit area.

These two-step optimization procedures show high accuracy and low variation at each step, so accuracy and deviation can be greatly improved compared to the method specified in the existing road design manual.

The following table 1 shows various attributes related to the size and operation of expressway rest areas, their classifications and data types for calculating the optimum size of expressway rest areas.

Variable type property Classification Data type Independent variable Main line Passing traffic volume () Large / day shame Highway name Gyeongbu, Namhae, Olympic, West Coast, Honam, Central, Tongyeong - Daejeon, Central Inland, Youngdong, Central designation direction Ascending, descending 〃 Crossing / branching direction Ascending, descending 〃 Location ratio 1 to 4 quartiles 〃 Nearby Promotion
Service area () km shame Nearby Ha
Service area () km 〃 service gas station Yes No designation Charging station Yes No 〃 Workshop Yes No 〃 Dependent variable Model 1 Usage Rate ()% shame Model 2 Revenue per vehicle () \ / 〃

As shown in Table 1, the independent variables that can describe the highway rest area are largely defined as lane property and service property. The independent variables of mainline property are passing traffic volume, number of expressway, direction of expressway, location from IC / LC, position ratio, Distance from upstream / downstream service area. The independent variables of the service attributes are set to the presence or absence of gasoline / LPG service station and repair shop.

The dependency variables that can be defined as functions of these independent variables are set as Usage rate of service area and Sales per vehicle.

Normally, there is only one dependent variable in the optimization problem, but in the present invention, since the optimization step is performed in two steps, a dependent variable to be optimized for each step is designated and two dependent variables are set.

In this case, it should be understood that the dependent variable of the first step is not a selected dependent variable to go through only the second step, and it is an advantage that it is a predictive value for calculating the convenience size of the optimal scale of the expressway rest area itself.

From the standpoint of these independent variables and dependent variables, the data obtained by examining the cases of the current highway rest area are stored in the case database. The various numerical values illustrated in the present specification are based on the results of an analysis by the present inventor of cases of 106 highway rest areas in Korea as of 2011. In contrast, the test case has independent variable values similar to the existing cases, but note that the values of dependent variables that are the subject of prediction are not yet available.

The following table 2 shows the correlation between some independent variables and dependent variables in the expressway rest area.

variable Passing traffic volume Usage Rate Service area sales Rest Area Construction Area Passing traffic volume One .862 .692 .383 Usage Rate .862 One .873 .435

As shown in Table 2, the correlation between the sales of the highway rest area and the passing traffic volume, the usage vehicle of the service area, and the building area of the service area Respectively.

In the existing road design manual, the traffic volume multiplied by the conversion factors is used to calculate the number of vehicles in the rest area and the parking area immediately. However, according to Table 2, the correlation coefficient between traffic volume and building area obtained from actual cases is only 0.383, and the correlation coefficient between the number of resting area vehicles and building area is only 0.435. This is in line with the fact that resting places built with parking lots and building areas according to current standards are difficult to operate.

On the other hand, the correlation coefficient between the passing traffic volume and the number of resting place vehicles is 0.862, which is highly correlated. This is because the predictive model 1 of the present invention for estimating the number of resting vehicle vehicles (that is, resting place utilization rate) Also proves valid.

However, the correlation coefficient between pass traffic volume and resting place sales is 0.692, which is not low, but it does not show enough correlation. Therefore, calculating the turnover of the rest area directly from the traffic volume may not achieve desirable results in terms of accuracy and variation.

On the other hand, the correlation coefficient between the number of resting place vehicles and the resting place sales is highly correlated with 0.873, which is the predictive model of the present invention estimating the resting place sales (ie, sales per vehicle) from the number of resting place vehicles 2 proves to be realistic.

It can be predicted that the resting place sales forecasted through the two-step estimation process of passing traffic volume → resting place utilization rate estimation → resting place sales estimation can be more realistic than resting place sales estimated directly from the passing traffic volume.

Predictive model 1 is a predictive model for predicting resting place utilization rate from passing traffic. The predicted resting place utilization rate and number of resting place vehicles in predictive model 1 can be used to calculate the size of facilities such as number of parking spaces, restrooms, have.

Predictive model 2 is a predictive model for estimating sales per vehicle (or resting place sales) in the number of resting places (or resting place utilization rate). Forecasted resting place sales in forecasting model 2 are the sum It can be used for

More specifically, as will be described later, based on the probability distribution curve of the sales per unit area obtained from the case analysis, the target sales per unit area is determined, and the sales per unit area is determined to achieve the forecast of the rest area sales It is possible to calculate the area of necessary operating facilities and operating facilities.

Thus, the total size of the expressway rest area can be calculated by summing the areas of convenience facilities, operating facilities, and operating facilities calculated from each prediction model.

FIG. 2 is a flowchart illustrating concrete steps of an optimal input resource calculating method of a test case project, which is an example of calculating an optimal scale of a highway rest area according to an embodiment of the present invention.

Referring to FIG. 2, in step S21, examples of the test case project including the target attribute as a target of prediction, for example, various properties that can define the characteristics of the expressway rest area in various aspects are specified , Analyzes the correlation between these various attributes, and selects an intermediate attribute having a high correlation with the target attribute among various attributes based on the analyzed correlation.

For example, the inventor has found that the problem of calculating the optimum size of a highway rest area is due to the problem of how much sales will occur at a highway rest area site where 10,000 cars pass each day and the sequential order of how large the rest area should be for such sales Lt; / RTI >

At this time, the rest area size can not be simply calculated as an area, but it is the sum of the sizes of facilities, sales facilities, and operating facilities with different characteristics. Convenience facilities do not directly contribute to sales due to parking lot, toilet, greenery, landscaping, but it is necessary to accommodate vehicles and passengers who use resting places during traffic. Since sales actually occur only at sales facilities such as kiosks and restaurants, sales at resting places can determine the size of sales facilities. The size of the facility depends on the size of the facility and the size of the facility.

In fact, as can be seen from the deficit management, the correlation between passing traffic volume and rest area size may not be high. The sales data per unit area collected in reality is similar to the gamma distribution curve, but the top 25% of the rest areas have annual sales of about 3 million won per 1 square meter of building area, while the bottom 25% Respectively. Therefore, there is no direct correlation between sales and size.

On the other hand, as previously indicated, the resting place sales have a sufficient correlation with the passing traffic volume.

Therefore, in view of this fact, it can be more accurate to set the target attribute as the target of prediction as the sales amount of the rest area rather than the rest area size.

The resting place utilization rate (or the number of resting place vehicles) can be selected as the intermediate property having a high correlation with the target rest area sales. The use rate of rest area is also a factor that can determine the size of convenience facilities in rest area, so it needs to be predicted as a kind of goal attribute itself in optimizing the optimal size of rest area of expressway.

In step S22, the middle attribute and the target attribute among the specific attributes in step S21 are set as first and second dependent variables, respectively, and the remaining attributes are set as independent variables, Builds a case database so that the cases of at least each of them are represented by at least independent variables and first and second dependent variables, and prepares a test case where test values of independent variables are given.

For example, the passing traffic volume and the like in Table 1 are independent variables, and the resting place utilization rate (or the number of resting vehicle vehicles) is the first dependent variable, and the sales per vehicle (or resting place sales) is the second dependent variable.

In this case, the actual cases of the case database may include independent variable values such as at least passing traffic data, and may include rest area utilization and dependent variable values of sales per vehicle. The test case includes at least independent test values such as passing traffic data, and can include resting place utilization and sales per vehicle as unknown dependent variables.

Throughout this specification, the resting place utilization rate is obtained by dividing the number of vehicles for resting place use by the amount of passing traffic per day, and the resting place utilization rate and resting place use vehicle number can be mixed according to the context with substantially the same meaning. Likewise, the sales per vehicle are divided by the number of resting place vehicles used for resting place sales, and they can be mixed according to the context with virtually the same meaning. Below, we will discuss the use of resting places and sales per vehicle.

Now we can predict the first and second dependent variables at each step by sequentially driving two-step prediction models based on case-based reasoning (CBR).

In step S23, in the first prediction model, the first dependent variable prediction value is estimated based on the result retrieved from the case database with respect to the independent variable test values of the given test case. The specific process of retrieving the case database will be described in detail later with reference to FIG. 3 to FIG.

Specifically, given the test values of the independent variables in the test case, the case database is first searched through the case-based reasoning methodology to extract similar cases.

The first dependent variable predicted value of the test case is then estimated from the first dependent variable values of the extracted similar cases. The estimate of the predicted value can be calculated as a statistical representative value of the first dependent variable values of the extracted similar cases, such as arithmetic mean, median, mode, and weighted average.

In the embodiment of estimating the size of a highway rest area, W resting place test cases are given predetermined independent variable test values as shown in Table 3, for example.

Project Properties data Name of resting place W Rest Area Passing traffic volume 12,031 Expressway number 10 times direction Ha line Direction from IC / JC Ascending Location ratio Second quartile (0.43) In-store resting street 29 km Haeam Rest Area 39 km gas station Yes Charging station no Workshop no

For example, the W resting place will be built on the downstream side of the Central Expressway with a traffic volume of 12,031 vehicles a day. The planned site will be equipped with a gasoline gas station only 29 km and 39 km away from the nearest up / down direction rest stop.

The similar cases retrieved from the case database through the case-based reasoning methodology to be described later in the first prediction model for these properties can be illustrated as shown in Table 4 below.

variable Case number Pass
Traffic volume
(Logarithm / day) Expressway number direction IC / JC
direction A resting place
Street
(km) Ha line
Rest area
Street
(km) gas station Charging station Workshop Rest area
Utilization rate
(%) vehicle
Use
Algebra
(Large / day) Test
case - 12,031 10 times Ha line Ascending 29.0 39.0 One 0 0 19.16 2,305 Discovered
Case 1 70 10,358 No. 6 Ha line Ascending 26.0 46.0 One One 0 19.34 Discovered
Case 2 25 13,103 No.2 Ha line Ha line 26.0 28.0 One One 0 18.98

The use rates of resting places in similar cases 1 and 2 were 19.34% and 18.98%, respectively.

Accordingly, the predicted value of rest area utilization rate of test cases can be calculated through statistical representative values such as arithmetic mean, median value, mode value, and weighted average value. When calculating by arithmetic mean, the predicted value of rest area utilization rate of test case can be estimated as 19.16%. The number of vehicles used at resting places in the day is calculated as 12,031 units * 19.16% = 2,305.

On the other hand, the rest area utilization rate prediction value is applied to the next step S24, and can be used for estimating the parking area for resting places, restrooms, green areas, and the like.

Subsequently, in step S24, in the second prediction model, based on the results of searching from the case database for the given independent variable test values and the first dependent variable predicted value estimated in step S23, . Likewise, the specific process of retrieving the case database will be described in detail later with reference to FIG. 3 to FIG.

In the embodiment of estimating the size of the highway rest area, W resting test case has a second dependent variable, for example, sales amount per vehicle, as shown in Table 3, with predetermined independent variable test values and estimated resting place utilization rate estimated as shown in Table 4 Is estimated.

The similar cases retrieved from the case database with the case-based reasoning methodology in the second prediction model can be illustrated as shown in Table 5 below.

variable Case number Pass
Traffic volume
(Logarithm / day) Expressway number direction IC / JC
direction A resting place
Street
(km) Ha line
Rest area
Street
(km) gas station Charging station Workshop Per vehicle
take
(Won / Large) Rest area
take
(Million / year) Test
case - 12,031 10 times Ha line Ascending 29.0 39.0 One 0 0 3,878 3,263 Discovered
Case 1 70 10,358 No. 6 Ha line Ascending 26.0 46.0 One One 0 4,104 Discovered
Case 2 97 10,966 10 times Ha line Ha line 30.0 36.0 One One One 4,247 Discovered
Case 3 102 7,601 10 times Ascending Ha line 29.7 83.0 One 0 0 3,284

From the sales values per vehicle for three similar cases, the forecasted sales per vehicle for the test case can be calculated through statistical representations such as arithmetic mean, median, mode, and weighted average. When calculated as an arithmetic average, the predicted sales value per vehicle of the test case can be estimated at 3,878 won /. In this case, annual resting place sales are calculated as 3,878 won * 2,305 units * 365 days = 3,263 million won.

The case-based reasoning methodology will be briefly described below. The case-based reasoning methodology applied in the present invention can be applied to other quantitative prediction methods that generally have high accuracy and low deviation to improve low accuracy and large deviation, which are common drawbacks of case-based reasoning, Based onference methodology (Hybrid CBR) that combines multiple regression analysis (MRA), artificial neural network (ANN), and genetic algorithm with case based reasoning.

In general, case-based reasoning (CBR) is a methodology that selects project cases with high similarity through comparative analysis of projects based on project characteristics. The core of the CBR methodology is the attribute similarity (AS), the attribute weight (AW), and the case similarity (CS). The case similarity CS between a real case and a test case can be quantified as the sum of the products of the property similarities AS and the property weights AW.

Where AS is the attribute similarity, AW is the attribute weight, CS is the case similarity, n is the number of attributes, and m is the number of cases.

The attribute similarity (AS) is based on the difference between the independent variables of the same kind between the actual and test cases, for each project characteristic, that is independent variables (which may in some cases also include dependent variables) . If the data of the independent variable is nominal, the attribute similarity is 1 if they are equal to each other, and 0 otherwise. If the data of the independent variable is numerical, the attribute similarity AS can be expressed by the following equation (2).

f _AS denotes the property similarity calculation formula, AV _{Test_Case} is the specific property value of the test case, and AV _{Retrieved_Case} is the specific property value of the retrieved case. f _AS is closer to 100% as the attributes are similar. The MCAS is a minimum criterion for scoring the attribute similarity, and the attribute similarity value must be valid if the attribute similarity is higher than MCAS, otherwise the value is discarded. The MCAS may be set to, for example, 10% and may be optimized as described below.

The similarity degree (CS) of a test case can be expressed by the following equation (3) by using the property similarity and attribute weight of all independent variables between the test case and the retrieval case, and how the test case is similar to each retrieval case.

f _CS is a function for calculating the case similarity (CS), f _AW is a function or the result for calculating the attribute weight (AW), f _AS is a function or the result of calculating the attribute similarity (AS) n is the number of attributes. The attribute similarity AS is calculated as shown in Equation 2, and the attribute weight AW can be empirically determined or mathematically optimized through other optimization methods. The case similarity (CS) can be illustrated as a value obtained by dividing the weighted property similarity obtained by adding the attribute similarity to the property weight by the sum of the property weights.

This general case-based reasoning (CBR) methodology provides similar cases with high degree of similarity in the case, and the prediction results can be calculated through statistical processing of similar cases. Therefore, the CBR methodology provides not only forecast results, but also similar cases to support them, so that we can explain how we obtained those forecasts. Accumulation of cases can also improve prediction accuracy.

However, the general CBR methodology generally tends to have poorer prediction accuracy than other quantitative prediction methodologies, for example, methodologies such as MRA (multiple regression analysis) and ANN (artificial neural network).

On the other hand, quantitative prediction methodologies such as MRA and ANN can not provide similar cases based on prediction accuracy.

In order to overcome this problem, the inventor has devised a hybrid CBR methodology using a combination of MRA, ANN, and GA (Genetic Algorithm) as follows. According to the hybrid CBR methodology devised in the present invention, both the high predictive accuracy of MRA and ANN and the merits of each methodology of providing a basis for similar cases of CBR can be provided.

FIG. 3 is a flowchart illustrating concrete steps of a hybrid case-based reasoning methodology applied in an optimal input resource calculation method of a test case project, which is an example of calculating an optimum size of a highway rest area according to an embodiment of the present invention.

3, at step S31, at least two quantitative prediction methodologies, e.g., MRA and ANN, are applied to m cases in the case database to determine the predicted values PV of the dependent variable .

In step S32, the error rate of the dependent variable prediction is calculated from the difference between the actual values AV of the dependent variable and the calculated predicted values PV in the m cases, for example, the average absolute error percentage MAPE : mean absolute percentage error) can be calculated by Equation (4).

Where f _MAPE is a function to obtain MAPE, AV is the actual value of the dependent variable, PV is the predicted value of the dependent variable, and m is the number of cases.

Since this error rate is based on a large number of actual cases, it is reasonable to assume that the same methodology will result in the same or similar error rate in new test cases. Also, since the error rate of the MRA or ANN methodology is usually smaller than the error rate of the CBR methodology, if the actual value of the case that is found to be similar by the CBR methodology is outside the range of the predicted value considering the error of the MRA or ANN methodology, It is desirable to remove the CBR case from the similar case candidates. In this specification, setting the filtering range based on these prediction ranges and eliminating the inadequate CBR case using this range is called filtering. The specific filtering operation is described in the following steps.

In step S33, each dependent variable predictive value is calculated for the test case based on at least two or more quantitative prediction methodologies, e.g., MRA and ANN.

In step S34, on the basis of the error rate MAPE calculated in step S32 and the predicted value PV of the dependent variable calculated in step S33, a prediction range PR).

Specifically, for example, the prediction range (PR _MRA ) in the _MRA methodology and the prediction range (PR _ANN ) in the ANN methodology can be defined as shown in Equation (5).

In Equation (5), although each prediction range PR is defined to have upper and lower limits by an average absolute error percentage (MAPE) around each predicted value PV, upper and lower limits are defined by different parameters according to the embodiment It is possible.

In step S35, a filtering range is determined based on at least two quantitative prediction methodologies, for example, a prediction range such as MRA and ANN. Here, the filtering range is applied to the filtering for the search cases to be performed in the later step S37.

Specifically, for example, the predicted range of the _MRA (PR _MRA ) and the predicted range of the _ANN (PR _ANN ) will be slightly different. The range in which the predicted ranges of the two methodologies overlap each other, that is, the cross- 6 and 7 as the filtering range.

Of the maximum of the upper limit value of the lower limit value of: (cross-range between the predicted value of the MRA and ANN models CRMA) is the prediction range of the MRA (PR _MRA) and the predicted range (PR _ANN) of the ANN mathematical filtering range of the equation (6) And may be illustratively specified as the Cross Range between the minimums.

The expanded filtering range (CRMA *) in Equation (7) can be illustratively specified as an expanded crossing range magnified by a tolerance range of CRMA for the filtering range (CRMA) in Equation (6).

This specified filtering range can be applied to cases retrieved through the CBR methodology for the test case in question.

Depending on the embodiment, in step S33, the prediction range itself of one CA estimation methodology may be used as a filtering range and applied to search cases.

In step S36, which may be optional, a plurality of parameters for at least case-based reasoning are optimized using a genetic algorithm.

4, FIG. 4 is a conceptual diagram illustrating a chromosome of a genetic algorithm (GA) in a method for calculating an optimum input resource of a test case project, which is an example of calculating an optimum size of a highway rest area according to an embodiment of the present invention .

In some embodiments, the chromosomes of the genetic algorithm include parameters for case-based reasoning such as MCAS (minimum criterion for scoring attribute similarity), RAWn (range of nth attribute weight), RCS (range of case selection) And may include parameters for determining the filtering range, for example, TRCRMA (Tolerance range of cross range between predicted values of MRA and ANN models).

For example, if the MCAS is too high, the degree of similarity may be lowered only in certain attributes, but cases with high similarity in other attributes may be dropped. For example, if the RAW is too high, the effect of certain attributes can be overestimated, and if the RCS is too small, there will be few cases to be searched and it may be difficult to calculate the predicted value of the dependent variable or present a past case. If the TRCRMA is too large, the prediction accuracy may be degraded.

Therefore, genetic algorithms can track changes in population over generations by selecting, breeding, mutating, and exchanging chromosomes with these parameters as the highest prediction accuracy. And chromosomes belonging to the one with the greatest number of populations.

Referring again to FIG. 3, in step S37, search cases are extracted through case-based reasoning. In step S38, only the search cases in which the dependent variable values are within the filtering range are filtered out It can be output as a similar case.

Depending on the embodiment, the case-based reasoning of step S37 may be performed without applying parameter optimization of step S36, for example, by applying empirical parameters.

On the other hand, the cases stored in the case database, for example, all have cases in common with cases of domestic highway rest areas, but they are linked to dense areas according to expressways (connecting between metropolitan areas and remote metropolitan areas or between adjacent provinces) The difference in the type of vehicle (lorry or passenger car) that is mainly used may cause a great difference in the usage rate of the rest stop or the sales per vehicle for each highway, for example.

Therefore, for more accurate similar case search, it may be desirable to exclude cases that may be searched as similar, though it is likely that they are indeed indifferent among the search cases.

According to an embodiment of the present invention, at the time of the case-based reasoning described above, the search target cases are divided into at least two clusters showing a significant difference in specific attributes, and the case is searched only in the cluster to which the test value of the independent variable belongs. Can be improved.

5 and 6, in an optimal input resource calculation method of a test case project, which is an example of calculating an optimum size of a highway rest area according to an embodiment of the present invention, FIG. 6 is a conceptual diagram illustrating a second cluster partition that divides a cluster of sales per vehicle according to a highway number. FIG.

Partitioning of clusters can be performed using various cluster analysis methods as a part of the data mining methodology. Among such cluster analysis methodologies, we use the Decision Tree (DT) methodology as an example for data classification.

Decision tree or decision tree methodology is a method that sequentially selects one of several independent variables that are statistically highly correlated with dependent variables and then sequentially classifies the dependent variable values with division criteria .

When applying the decryption tree methodology, algorithms such as Classification and Regression Trees (CART), chi-squared automatic interaction detection (CHAID), and C4.5 are used. The CART technique forms a binary tree arti- cle model, which is a simple method to divide the independent variables that best divide the dependent variable and its subdivisions into two groups . In this case, a classification tree is used if the dependent variable is a nominal value, and a regression tree is used if it is a ratio value. In the example of FIGS. 5 and 6, the CART technique is used.

As shown in Table 2, the use rate of resting place, which is a dependent variable to divide the cluster, shows a high correlation coefficient with the passing traffic volume. Experience shows that the average speed is reduced when there is a lot of traffic, so the proportion of vehicles using rest areas will be somewhat reduced in order to save running time. Therefore, in this case, the resting place utilization rate can be divided on the basis of the passing traffic amount of a specific value as a dividing reference.

Referring to FIG. 5, cluster partitioning of the case database for prediction model 1 is illustrated.

The average of the resting place utilization rate was 0.211 in 106 cases before split. In case of passing traffic volume less than or equal to 16,521 vehicles / day, 61 cases with pass traffic volume less than 16,521 and average resting place utilization rate 0.246 And 45 communities with transit traffic volume greater than 16,521 and average resting place utilization rate of 0.165.

This time, the revenue per dependent variable vehicle can be divided into highway numbers because it will be related to the normal usage purpose of the highway. For example, if a freeway is frequently used on a freeway, it is expected that the number of vehicles that use transit traffic or rest area is lower than that of buses or passenger cars. Because the nature of such a highway can be represented by a highway number, it is possible to divide the sales per vehicle by the division number of the highway number.

Referring to FIG. 6, cluster partitioning of the case database for prediction model 2 is illustrated.

In a total of 106 cases before the split, the average of the sales per vehicle was 1,398 won / b / day. When the division number of the expressway number is 1, 2, 3, 5, 6, 8 and 4, 7, 9, The third group consists of 64 cases with highway numbers 1, 2, 3, 5, 6, and 8, respectively, and the average sales per vehicle is 1,205 won / vehicle / day. , 9, and 10, and 42 cases with average sales per vehicle of 1,694 won / vehicle / day can be generated.

By dividing this into clusters, similar cases are searched among the relevant cases in each prediction model, so that the accuracy of the case search can be improved and the deviation can be lowered.

The cluster analysis methodology for the case-based reasoning (CBR) methodology or the hybrid case-based reasoning methodology described above with reference to FIGS. 3 to 6, respectively, can be selectively applied according to the embodiment.

Returning to FIG. 2, in the optional step (S25), when the second dependent variable value is related to performance, a probability distribution curve of the performance according to the input resource is derived from the cases stored in the case database, To determine the amount of resources to be put in order to achieve the second dependent variable predicted value in the test case project with the selected target performance.

Specifically, for example, in a case of a highway rest area, the second dependent variable, the resting place sales, can be analyzed based on the resource of the building area, and the probability distribution function of the per-unit sales per unit area can be derived.

7 is a graph illustrating a probability distribution function for achieving sales per unit area in an optimal input resource calculating method of a test case project, which is an example of calculating an optimum size of a highway rest area according to an embodiment of the present invention. to be.

The graph of FIG. 7 shows an empirical probability of achieving sales per unit area, which is similar to the gamma distribution function. The number on the graph is the percentile. For example, a resting place with 75% below (ie, the top 25%) means that it can raise sales of 2,473,000 won per unit area per year.

Next, when the decision maker sees the probability distribution curve and selects the target performance, it can determine the amount of resources to be put in order to achieve the second dependent variable predicted value with the selected target performance.

For example, in the case of a test case project such as W rest area in Table 3, a decision maker can select a target performance such as sales per unit area by referring to the empirical probability obtained from the existing cases as shown in FIG.

For example, decision makers can be extremely conservative, aiming for only 25 per cent of sales per unit of floor space, 50 per cent for middle or average, or boldly targeting the top 25 per cent.

Table 6 is a table showing the building area of the sales facility (including the operation facility) required to achieve the resting place sales forecast value according to each target performance. In the first place, suppose that the decision maker designed the rest area with a building area of 1,100 pyong, or about 3,663 ㎡, based on the current regulations, without optimizing the rest area size.

Alternatives Sales per unit area
(Thousand won / year / ㎡) Forecasted rest area
Optimal area
(㎡) Actual design
Area of rest area
(㎡) Difference (㎡) Alternative 1 Percentile 25% 1.127 2,896.15 3,663 + 767 Alternative 2 Percentile 50% 1.700 1,919.86 + 1,744 Alternative 3 Average 1.917 1,702.19 + 1,961 Alternative 4 Percentile 75% 2.473 1,319.64 + 2,344

Referring to Table 6, Alternative 1 aims to achieve sales per unit area of the lower 25%, which can be estimated with the estimated construction area of 2,900 ㎡. This means that sales forecasts can be achieved only when 767 ㎡ is less than the actual designed rest area. In the case of Alternative 4, which has selected relatively bold target performance, it is estimated that the rest area sales value can be achieved with a building area of 1,300 ㎡. Currently, the rest area design area exceeds 2,344 ㎡, which is about 3 It means that it is overly planned.

W Unless the number of vehicles used suddenly increases suddenly due to other reasons at resting places, actual sales will not exceed the sales forecast value even if the rest area is opened beyond the optimum area, and the excess area will not contribute to sales but costs only. . ≪ / RTI >

The present invention can eliminate the causes of deficits and calculate the optimal facility size and optimal sales facility scale at a desired location.

Furthermore, by repeatedly executing the optimum size calculation method of the present invention while slightly changing the information about the location of the site along the highway route, it is possible to find a point where the sales amount is expected to be the largest under the condition of observing the regulations.

FIG. 8 is a block diagram illustrating an optimal input resource calculating apparatus of a test case project illustrating an optimum scale calculation of a highway rest area according to an embodiment of the present invention.

Referring to FIG. 8, the optimal input resource calculating device 80 is an apparatus for implementing the optimum size calculating method of the present invention described with reference to FIGS. 1 to 7, and includes a case database 81, a first cluster dividing unit 82 A first prediction model 83, a second grouping division 84, a second prediction model 85, a resource-performance probability distribution analyzing section 86, and an optimal input resource determining section 87 have.

The case database 81 stores instances represented by independent variables and values of dependent variables from existing executed projects.

The first cluster partitioning unit 82 and the second cluster partitioning unit 84 analyze the cases in the case database 81 based on each of the specific partitioning independent variables having a relatively high correlation with each dependent variable And divided into at least two clusters. The respective clusters required by the first or second predictor 83 or 84 among the clusters divided by the first and second cluster dividers 82 and 84 are divided into first and second predictors 83 and 84 ).

Among the cases belonging to the cluster to which the independent variable test value belonging to the division is based among the cases of the case database 81 or the cluster provided by the first cluster division unit 82, We search for cases similar to the test case with given independent variable test values and estimate the first dependent variable predictive value for the test case based on the first dependent variable values of the retrieved similar cases.

Likewise, the second predictor 85 may be a case that belongs to the cluster to which the independent variable test value belonging to the division criterion among the cases of the case database 81 or the cluster provided by the second cluster division unit 84 is divided Among the plurality of test cases, a case similar to the test case having the given independent variable test values and the first dependent variable predictive values, and estimates the second dependent variable predictive value for the test case based on the second dependent variable values of the retrieved similar cases .

The first and second prediction units 83 and 85 include quantitative prediction analysis units 831 and 851 and a case based reasoning models 832 and 852 that respectively perform a plurality of different quantitative prediction methodologies, , Filtering units 833 and 853 for filtering the search results of the case-based reasoning models 832 and 852, a case-based reasoning models 832 and 832 and a filtering unit 830 for filtering the search results of the case-based reasoning models 832 and 852 with a filtering range determined from the error rates and prediction ranges of the respective quantitative prediction methodologies And / or parameter optimizing units 834 and 854 that optimize the parameters of the filtering units 833 and 853, respectively.

The resource-performance probability distribution analyzing unit 86 analyzes the empirical probability distribution on the performance per input resource based on the cases of the case database 81, and can provide it to the decision maker.

The optimal input resource decision unit 87 can analyze the resources that a decision maker has to input in order to achieve the value of the dependent variable predictions with a certain target performance based on the empirical probability distribution and provide this to the decision maker.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, but, on the contrary, Modification is possible. Accordingly, the spirit of the present invention should be understood only in accordance with the following claims, and all of the equivalent or equivalent variations will fall within the scope of the present invention.

Further, the apparatus according to the present invention can be implemented as a computer-readable code on a computer-readable recording medium. A computer-readable recording medium includes all kinds of recording apparatuses in which data that can be read by a computer system is stored. Examples of the recording medium include a ROM, a RAM, an optical disk, a magnetic tape, a floppy disk, a hard disk, a nonvolatile memory, and the like, and a carrier wave (for example, transmission via the Internet). The computer-readable recording medium may also be distributed over a networked computer system so that computer readable code can be stored and executed in a distributed manner.

80 Optimal input resource calculation device 81 Case database
82 first cluster division part 83 first prediction section
84 second group division part 85 second prediction part
86 resource-performance probability distribution analyzing section 87 Optimum input resource determining section

Claims (24)

As a method for calculating the optimal size of a highway rest area using a computer,
The computer comprising:
Providing a test case in which each case is represented by a plurality of independent variables including at least a passing traffic volume, a case database represented as a dependent variable including resting place utilization and sales per vehicle, and a test case given independent variable test values;
Estimating a rest area utilization rate prediction value based on a result retrieved from the case database with respect to the given independent variable test values;
Estimating a sales forecast value per vehicle based on a result retrieved from the case database regarding the given independent variable test values and the rest area utilization predicted value; And
And determining a scale of a rest area facility capable of achieving a target service rate from the rest area utilization rate predicted value. The method according to claim 1, wherein estimating the resting place utilization rate prediction value comprises:
Dividing the resting place utilization rate values of the cases in the case database into at least two clusters according to the passing traffic criterion;
Searching for similar cases with respect to the given independent variable test values in a cluster to which the given passing traffic belongs; And
And estimating a rest area utilization rate prediction value by referring to rest area utilization values of the retrieved similar cases. 4. The method of claim 1, wherein estimating the per-vehicle sales forecast value comprises:
Dividing the sales values per vehicle of the cases in the case database into at least two clusters according to a highway number reference;
Retrieving similar cases with respect to the independent variable test values and the rest area utilization predictive value given in the community to which the highway to which the rest area is to be installed belongs; And
And estimating a sales forecast value per vehicle by referring to sales values per vehicle of the searched similar cases. The method of any one of claims 1 to 3,
Determining a filtering range based on respective dependent variable predictions and error rates based on at least two quantitative prediction methodologies;
Optimizing at least parameters for case based reasoning using a Gene Algorithm; And
And outputting, as a similar case, search cases in which dependent variable values are within the filtering range, among search cases obtained from case-based reasoning based on optimized parameters. 5. The method of claim 4, wherein the at least two quantitative prediction methodologies include an MRA (Multiple Regression Analysis) and an ANN (Artificial Neural Network). 5. The method of claim 4,
Range between a maximum of the lower bounds of the prediction ranges obtained from the at least two prediction methodologies and a minimum of the upper bounds of the prediction ranges obtained from the at least two prediction methodologies. A method for calculating the optimum size of a. 5. The method of claim 4,
An intersection range between a maximum of the lower bounds of the prediction ranges obtained from the at least two prediction methodologies and a minimum of the upper bounds of the prediction ranges obtained from the at least two prediction methodologies by an allowance tolerance range Wherein said method comprises the steps of: 8. The method of claim 7, wherein the at least case-
As the parameters for case-based reasoning, MCAS (Range Criteria for Scoring Attribute Similarity), RAWn (Range of Case Attribute Weight), RCS (range of case selection) and parameters for determining the filtering range, Tolerance range of cross range between the predicted values of MRA and ANN models). The method according to claim 2 or 3, wherein the partitioning of the cluster is performed according to a cluster analysis methodology including a decision tree methodology. The method according to claim 1,
If the sales per unit area is selected based on the probability distribution function of the sales per unit area derived from the cases stored in the case database, the sales facility size of the rest area where the sales value pertaining to the rest area sales can be achieved with the sales per unit area Further comprising the step of determining the optimal size of the highway rest area. The method according to claim 1,
Further comprising the step of estimating an area of a convenience store including a parking lot of a highway rest area based on the predicted value of the rest area utilization rate. As an optimal input resource calculation method of a computer-based test case project,
The computer comprising:
Providing a test case in which each case is represented by a plurality of independent variables and first and second target variables, and an independent variable test value;
Estimating a first dependent variable predictive value based on a result retrieved from the case database with respect to the given independent variable test values; And
Estimating a second dependent variable predictive value based on a result retrieved from the case database with respect to the given independent variable test values and the first dependent variable predictive value,
The independent variables correspond to each of the attributes given test values for the test case among the properties of past cases stored in the case database for case based reasoning,
Wherein said dependent variables correspond to each of the properties whose predictions are to be estimated with respect to the test case through case based reasoning. The method of claim 12, wherein estimating the first dependent variable predictive value comprises:
Dividing the first dependent variable values of the cases in the case database into at least two clusters according to a first split criterion independent variable;
Searching for similar cases with respect to the given independent variable test values in the cluster to which the test value of the first split criterion independent variable belongs; And
And estimating a first dependent variable predicted value by referring to the first dependent variable values of the retrieved similar cases. The method of claim 12,
Wherein estimating the second dependent variable predicted value comprises:
Dividing the second dependent variable values of the cases in the case database into at least two clusters according to a second split criterion independent variable;
Searching similar cases for the given independent variable test values and the first dependent variable predicted value in a cluster to which the test value of the given second split criterion independent variable belongs; And
And estimating a second dependent variable predicted value by referring to second dependent variable values of the searched similar cases. The method of any one of claims 12 to 14,
Determining a filtering range based on respective dependent variable predictions and error rates based on at least two quantitative prediction methodologies;
Optimizing parameters for case based reasoning using a Gene Algorithm; And
And outputting, as a similar case, search cases in which dependent variable values are within the filtering range, among search cases obtained from case-based reasoning based on optimized parameters. 16. The method of claim 15, wherein the at least two quantitative prediction methodologies include an MRA (Multiple Regression Analysis) and an ANN (Artificial Neural Network). 16. The method of claim 15,
A cross-range between a maximum of the lower bounds of the prediction ranges obtained from the at least two prediction methodologies and a minimum of the upper bounds of the prediction ranges obtained from the at least two prediction methodologies. A method for calculating the optimum input resource of a project. 16. The method of claim 15,
An enlarged intersection range in which an intersection range between a maximum of the lower limit values of the prediction ranges obtained from the at least two prediction methodologies and a minimum of the upper limit values of the prediction ranges obtained from the at least two prediction methodologies is enlarged by an allowable marginal rate. A method for calculating the optimal input resource of a test case project. The method of claim 13 or claim 14, wherein the partitioning of the cluster is performed according to a cluster analysis methodology including a decision tree methodology. The method according to any one of claims 12 to 14,
Deriving a probability distribution curve of performance based on the input resource from the cases stored in the case database when the second dependent variable value is related to the performance, Further comprising the step of determining an amount of input resources to be input to achieve the second dependent variable predicted value in the project. A case database in which each instance is represented by a plurality of independent variables and first and second dependent variables;
A first predictor for estimating a first dependent variable predictive value based on a result retrieved from the case database with respect to the given independent variable test values, when a test case given with independent variable test values is provided; And
And a second predictor for estimating a second dependent variable predicted value based on a result retrieved from the case database with respect to the given independent variable test values and the first dependent variable predicted value,
The independent variables correspond to each of the attributes given test values for the test case among the properties of past cases stored in the case database for case based reasoning,
Wherein said dependent variables correspond to each of the properties whose predictions are to be estimated with respect to the test case through case-based reasoning. 23. The method of claim 21,
And dividing the dependent variable values of the cases in the case database into at least two clusters according to the split criterion independent variable and providing the divided clusters to the first predictor or the second predictor. Optimal input resource calculation device of project. 22. The apparatus of claim 21, wherein each of the first predictor or the second predictor comprises:
A quantitative prediction analyzer for computing dependent variable predictive values, error rates and prediction ranges with respect to the cases of the case database based on at least two quantitative prediction methodologies;
A case based reasoning model for searching cases similar to test cases using case based reasoning among the cases of the case database;
A parameter optimizer for optimizing parameters for the case based reasoning model using a genetic algorithm; And
And a filtering unit for filtering the cases retrieved from the case-based reasoning model with a filtering range determined based on the dependent variable predictive values, error rates and prediction ranges, and outputting a similar case. Resource calculating device. 23. The method of claim 21,
A resource-performance probability distribution analyzing unit for analyzing an empirical probability distribution on the performance per input resource based on the cases of the case database; And
And an optimal input resource determining unit for analyzing a resource to be input by the decision maker to achieve the second dependent variable predicted value with a specific target performance based on the empirical probability distribution. Resource calculating device.

Images