KR20110108127A - Cost estimating model and method for building projects in planning phase using cbr and ga - Google Patents

Abstract

The present invention relates to a construction cost prediction model for calculating a project budget at the stage of planning a new project of a building (especially a public building).
The construction cost prediction model according to the present invention collects case data from the completed ready-made project, selects optimal similar cases using case based reasoning (CBR) and genetic algorithm (GA), and similar cases. It is characterized by being configured to predict the construction cost of a new project based on the construction cost information.

Description

Cost estimating model and method for building projects in planning phase using CBR and GA}

The present invention relates to a system and method for estimating construction costs for easy project budgeting at the stage of planning a new project of a building (especially a public building). (Case Based Reasoning, CBR) and Genetic Algorithm (GA) are used to select the optimal similar cases and to construct the system and method for predicting the construction cost of a new project based on the construction cost information of similar cases. .

The new project planning phase of the building is to establish a business plan. The feasibility review is conducted based on project-related data surveys, and the project size, total project cost, and project period are determined by referring to the case of similar projects.

The total project cost of the building is composed of construction cost, compensation cost, and facility unit cost (design cost, supervision cost, and facility cost). Reasonable total project cost calculation through accurate calculation of cost is very important in establishing the project budget. In the current project of a new building, the construction cost calculation method in the planning stage is a method of multiplying the construction cost per floor area of similar cases by the estimated total floor area of the building to be constructed.

The planning stage does not come up with a specific design plan. Since the cost forecast depends on the cost information of the existing similar cases, the selection of similar cases similar to the new project greatly affects the accuracy of the cost forecast.

However, the current construction cost calculation may be based on the experience of the person in charge of the selection of similar cases, the inappropriate similar case may be selected, and the calculation of the construction cost is also based on the total construction cost, which is likely to reduce the accuracy of the forecast construction cost.

The present invention has been developed to solve the problem of low accuracy in the conventional construction cost prediction, has the following technical problems.

First, we will provide a construction cost forecasting model that can improve the accuracy of construction cost forecasting by selecting the optimal similar cases.

Second, we will provide a construction cost prediction model for the model that can be predicted by dividing the construction cost by the construction cost by construction type such as frame construction cost, finishing construction cost, and construction cost.

In order to solve the above technical problem, the present invention collects existing construction cost data, constructs a case database, selects the best analogous case using case-based reasoning (CBR) and genetic algorithm (GA) methodology, and then constructs the construction cost. It provides a cost estimating model that can estimate the detailed construction cost by dividing it into the details of cost, finishing construction cost and construction cost.

According to the present invention, the following effects can be expected.

First, by applying both case-based reasoning and genetic algorithm at the same time, it is possible to select the best analogous case and thus improve the accuracy in the construction cost prediction of the building. As a result, it is possible to establish goals and strategies appropriate to all situations at the planning stage of the building.

Second, it is possible to select the optimal similar case for each new project for the new project by reflecting the difference in the factors of construction cost for each construction type, and thus it is possible to more accurately predict the construction cost for each construction type.

Third, it can be usefully used for the construction cost prediction of buildings that can easily quantify the design variables that affect the construction cost calculation, such as public buildings.

1 shows a framework for a construction cost prediction model according to the present invention.
2A is a block diagram of a construction cost prediction model according to the present invention.
2b shows a process for estimating attribution weighting and attribution similarity criteria in a construction cost prediction model according to the present invention.
3 to 6 show screens executed when the construction cost prediction system according to the present invention is implemented on a computer.

It will be described below according to the accompanying drawings and preferred embodiments.

1 shows a framework for a construction cost prediction model according to the present invention. The construction cost prediction model according to the present invention has a basic feature in that it uses 'CBR (case-based reasoning) + GA (gene algorithm)' as an engine.

 The CBR methodology calculates case similarity scores by calculating and adding similarity scores between new cases and attributes of existing cases (design information), and selects cases with the highest case similarity scores as best similar cases. At this time, since the importance of each attribute is different in determining similar cases, weights are assigned to each attribute similarity, and it is not easy to calculate such weight. To solve this problem, we applied GA methodology. GA methodology derives the optimal value for each factor by setting the factors affecting the target variable to vary within a certain range.

As such, the present invention is based on case-based reasoning, but improves the accuracy in the selection of optimal similar cases by using the genetic algorithm together. Accordingly, it is expected that the accuracy of construction cost forecast will be improved.

Figure 2a is a block diagram of a construction cost prediction model according to the present invention, Figure 2b shows a process for calculating the attribute weight value and attribution similarity criteria in the construction cost prediction model according to the present invention, Figures 3 to 6 is Figure 2a And a screen realized when the model according to FIG. 2B is implemented. The construction cost prediction model according to the present invention is a model in which a series of steps of constructing a case database, inputting project information, evaluating similarity, generating similar cases, and estimating construction cost are implemented in a computer. Accordingly, the present invention distinguishes and proposes a system invention composed of a data processing module and a method invention composed of a data processing method. However, since the method invention may be referred to as a process of performing the system invention step by step, the following describes the detailed configuration of the present invention by tying together the system invention and the method invention, and also exemplifies the process of predicting the construction cost of public buildings.

(1) Case database construction

Since the present invention is based on CBR (case-based reasoning), a case database for a ready-made project of a completed building should be constructed, and in the present invention, the case database includes design summary information and cost information, which are factors influencing the construction cost calculation. It is constructed to be stored as information.

3 is an example of constructing a case database for a public project of a public building, and the design outline information includes the type of building, function, area, land area, total floor area, construction rate, floor area, landscaping area, public announcement, parking lot, and groundwater. , Information on ground floor number, eco-friendly grade, structural type, exterior window type, exterior material type, information and communication grade, and construction cost information includes direct construction cost (framework cost, finishing cost, non-building cost) by construction type (machinery, electricity, communication, Civil engineering, landscaping) and overhead costs. However, the design overview information has essential influence factors that are essential to the cost calculation regardless of the construction type by the basic design conditions, while selectively affecting the construction cost calculation by the addition of special design conditions or the difference in construction type. Since there may be a selection influencing factor, in consideration of such circumstances, in the present invention, the design overview information is divided into essential influence factors and selection influence factors for each type of work as shown in Table 1 below. Essential influencers are design outline information that is collected in the process of building case database, and optional influencers are design outline information that is additionally collected only when applicable.

[Table 1] Classification of Influence Factors of Design Outline Information

On the other hand, the representation of the design overview information can be divided into numerical and nominal variables as shown in [Table 2], the numerical value in the process of calculating the attribute similarity according to CBR (case-based reasoning) The nominal variable should be appropriately numerically applied. [Table 3] shows an example of digitizing the nominal variable of [Table 1].

TABLE 2

Classification of Design Overview Information by Expression Method

[Table 3] Transformation Values for Numerization of Design Outline Information Expressed as Nominal Variables

(2) Project information input module / step

In the project information input module / stage, design overview information about a new project of a building is received. 4A shows a design overview information input window for a new project of a public office, and may be configured to be inputted with the same items as the design overview information for a ready-made project. In other words, in the design outline information in [Table 1], properties belonging to essential influence factors are composed as basic information window so that they can be inputted. Additional properties window (system user inputs) can be selectively inputted. Input screen appears only when desired). FIG. 4B shows a registration window in which design outline information of a new project for calculating a construction cost is input as input through the input window of FIG. 4A.

(3) Similarity Evaluation Module / Step

In the similarity evaluation module / stage, the case similarity of the ready-made project for the new project is calculated by linking the design overview information of the new project inputted through the project information input module with the design overview information of the ready-made project stored in the case database. The similarity evaluation module / step consists of the property weighting calculation government / step, the property similarity failure government / step, the case similarity failure government / step, and the upper level extractor / step, the attribution weighting compensation government / step May be configured to further include, and furthermore, the property likelihood miscalculation / step may be composed of a reference setting part / step and an operation part / step.

① Attribute Weighting Government / Step

In the attribute weight calculation unit / step, as shown in FIG. 2B, a genetic algorithm using a combination of attribute weight values of the design outline information as a generation is performed on various pieces of information on the existing project stored in the case database. Optimize. In other words, the weight of each factor of influence of the design overview information is calculated by the optimization method by genetic algorithm.

As shown in [Table 1], the design overview information that affects each model may be different. In this case, the combination of attribute weights of the design overview information may be optimized for each model. In other words, the attribute weighting value for the construction cost calculation is the design overview information that is the influencing factor of the construction cost (Table 1 shows the building type, function, area, land area, total floor area, construction rate, floor area, landscaping area, public announcement, parking lot, Only the basement floor, the ground floor number, and the structural form) optimize and estimate the combination of attribute weighting values, and the same for the finishing and non-building costs.

② Property-Like Dosan Government / Stage

In the attribute similarity miscalculation / stage, the design similarity information of the new project and the design overview information of the existing project are linked to each other, and the property similarity diagram of the ready-made project for the new project is calculated by the following [Equation 1]. In the following Equation 1, the MCAS may be set by receiving an arbitrary value by a system user or automatically receiving an optimized value by a genetic algorithm.

[Equation 1]

here, f _AS (x) = property similarity estimation function

AV _{Test _ Case} = Design Outline Information by Property of New Project (Value)

AV _{Retrieved _ Case} = Design Outline Information (Value) by Property of the Original Project

MCAS = set attribute similarity criterion (0 ~ 1)

For example, if the property similarity of the total area defined by the ratio scale is calculated according to [Equation 1] above, the property similarity is 0.83 when the new project (Test_Case) has a total surface area of 100 and the existing project (Retrieved_Case) has a total floor area of 100. If MCAS is set to 0.83 or less, this attribute similarity value will be a valid value. If MCAS is set to more than 0.83, this attribute similarity value is not recognized and is given as 0.

If the MCAS is set as a method of optimizing by genetic algorithm in [Equation 1], the attribute likelihood calculation / step may be composed of a reference setting part / step and an operation part / step. In the reference setting section / step, as shown in FIG. 2B, the genetic similarity of the design overview information is executed by executing the genetic algorithm based on the property similarity criterion (MCAS) of the design overview information, which is stored in the case database. Optimize the criteria (MCAS). In this case, if the generation for executing the genetic algorithm is set to the attribute similarity criterion (MCAS) combination of the design outline information, the attribute similarity criterion (MCAS) combination will be optimized. In the calculation section / step, the function of [Equation 1] is stored, and the design overview information for each property of the new project and the ready-made project is calculated along with the property similarity criterion (MCAS) set through the reference setting section. As a result, attribute similarity is calculated.

③ Case of similar insolvency government / stage

In the case similarity mountainous government / stage, the property similarity value calculated through the property weighted mountain government and the property similarity calculated through the property similarity mountain government are calculated to calculate the case similarity of the established project as [Equation 2]. That is, the case similarity is scored by dividing the cumulative sum of the weighted attribute similarity (attribute weight × attribute similarity) by the cumulative sum of the attribute weight.

[Equation 2]

Where f _CS (x) = case similarity estimation function)

f _ASi = Attribute Similarity _{Diagram by} Design Outline Information

f _AWi = Attribute Weight by Design Outline

For example, if the case similarity for calculating the construction cost is calculated according to the above [Equation 2] as follows. If the property similarity of the total area is 0.83 and the property weight of the total area is 0.72, the weighted attribute similarity is 0.6225 (= 0.83 × 0.72) for the total area. After obtaining the cumulative sum of and dividing by the cumulative sum of the attribute weights, the case similarity is calculated.

④ Higher level extractor / step, attribute weight value compensation / step, case-like protection / step

In the upper level extractor / step, the case database of case similarity calculated through the case similarity bank / step is extracted from the case database. In the attribute weight correction / step, a combination of attribute weights of the design overview information is executed by executing a genetic algorithm that combines the attribute weight combinations of the design outline information into generations, targeting various information on the existing project extracted through the upper level extractor / step. To optimize. In the case similarity preservation government / stage, the case similarity degree of the established project is corrected to [Equation 2] by receiving the attribute similarity calculated through the property similarity misprediction / step and the attribute weight value corrected through the attribute weight correction / step.

(4) Similarity Generation Module / Step

The similar case generation module / stage integrates various information about the existing project stored in the case database with the case similarity information calculated through the similarity evaluation module, and integrates the similarity of the old project including construction cost information and case similarity information per floor area of the old project. Generate information. If the construction cost information stored in the case database is subdivided into direct construction cost information by sector, the similar case generation module / stage generates similarity integration information including direct construction cost information and case similarity information per floor area by construction type. In case of further inclusion, similarity integrated information is further generated including indirect construction cost information.

Figure 5a is a screen showing the similarity integrated information generated for the construction cost calculation in the construction cost calculation for the new project of public buildings, Figure 5b is a screen showing the similarity integration information generated for the calculation of the closing construction cost. 5a and 5b (case) similarity shows different values, because the attribute weighting value has a different value according to the type of work (frame, finish) influence factors in the calculation of case similarity.

If the similarity evaluation module / step further includes a higher level extractor / step, the similar case generation module / step is corrected through various information on the existing project extracted through the upper level extractor / step and the case similarity calculation. Similarity similarity information is generated from the similarity information of the existing projects. If the higher level extractor / step in the similarity evaluation module / step is further included along with the attribute weight correction / step and the case similarity compensation / step, the similar case generation module / step is extracted through the higher level extractor / step. The similarity integrated information of the ready-made project is generated from the information on the ready-made project and the case-likeness information corrected through the case similarity security / step. As a result, only similar cases with high case similarity are selected.

(5) Construction Estimation Module / Step

In the construction cost estimation module / step, the construction cost of the new project is calculated by receiving the construction cost information per floor area of the previous project from the similarity integrated information of the existing project generated through the similar case generation module / step. The construction cost information per floor area of the existing project may be the construction cost information per floor area of the best similar cases with the highest case similarity, or may be the construction cost information per average total floor area of the similar case groups with the highest case similarity. The construction cost information per floor area of the received ready-made project may be configured to be modified and input by the system user.

If the construction cost information stored in the case database is subdivided into direct construction cost information by construction type, the construction cost estimation module / stage receives direct construction cost information per floor area of the established project and calculates the construction cost of the new project as the construction cost by construction type. If it includes more indirect cost information, the indirect cost is calculated to be included.

Figure 6 shows the construction cost calculation window for the new project of the public office, as shown, it is configured to calculate the construction cost by receiving the construction cost information per floor area of the existing project as it is linked to the similarity integrated information.

The present invention has been described in detail above with reference to specific embodiments, but the embodiments are only for illustrating the present invention, and thus the embodiments substituted, added, and modified within the scope without departing from the spirit of the present invention are also described below. It will be said to belong to the protection scope of the present invention as defined by the claims appended hereto.

Claims (10)

As a system for estimating and predicting construction costs at the stage of planning a new project for a building,
A case database constructed to include design outline information and construction cost information which are factors influencing the construction cost calculation for the ready-made project of the completed building;
A project information input module for receiving design overview information about a new project of a building;
A similarity evaluation module for estimating the case similarity of the ready-made project for the new project by linking the design outline information of the new project inputted through the project information input module with the design outline information of the ready-made project stored in the case database;
Similar case generation is generated by integrating the various information about the existing project stored in the case database and the case similarity information calculated by the similarity evaluation module to generate the similarity integrated information of the old project including construction cost information and case similarity information of the existing project. module;
The construction cost estimation module for calculating the construction cost of a new project by receiving the construction cost information per floor area of the existing project from the similarity integrated information of the existing project generated through the similar case generation module;
The similarity evaluation module,
An attribute weight calculation unit for optimizing the attribute weight combinations of the design summary information by executing a genetic algorithm with the generation of the attribute weight combinations of the design summary information for various information on the existing projects stored in the case database;
Attribute similarity calculation which calculates attribute similarity of the ready-made project for the new project by the property of the design-information information by linking the design summary information of the new project and the design overview information of the ready-made project;
Case-like similarity government that calculates the case similarity of the established project by receiving the optimized attribute weight value and the attribute similarity calculated through the property similarity calculation government [Equation 2];
Construction cost prediction system of a building characterized by consisting of.
[Formula 1]

here, f _AS (x) = property similarity estimation function
AV _{Test _ Case} = Design Outline Information by Property of New Project (Value)
AV _{Retrieved _ Case} = Design Outline Information (Value) by Property of the Original Project
MCAS = set attribute similarity criterion (0 ~ 1)
[Formula 2]

Where f _CS (x) = case similarity estimation function
f _ASi = Attribute Similarity _{Diagram by} Design Outline Information
f _AWi = Attribute Weight by Design Outline In claim 1,
The similarity evaluation module,
And a high level extracting unit for extracting various kinds of information of the ready-made project belonging to the upper level, the case similarity calculated through the case similarity mountain government in the case database.
The similar case generation module,
A construction cost forecasting system for a building, characterized in that it is configured to generate similarity integrated information of a ready-made project with information on the ready-made project extracted through a high level extractor and case-likeness information calculated by the case-like property government. In claim 1,
The similarity evaluation module,
A high level extracting unit for extracting various kinds of information of the existing projects belonging to a higher level, the case similarity calculated through the case similarity mountain government in the case database;
An attribute weighting compensator for optimizing the attribute weight combination of the design summary information by executing a genetic algorithm with the generation of the attribute weight combinations of the design outline information for the various information on the existing project extracted through the upper level extraction unit;
And the case similarity compensation government that corrects the case similarity of the established project by [Equation 2] by receiving the attribute similarity calculated through the attribute similarity mountain government and the attribute weight value corrected through the attribute weight value correction government.
The similar case generation module,
A construction cost prediction system for a building, characterized in that it is configured to generate similarity integrated information of a ready-made project with various information on the ready-made project extracted through a high level extraction unit and case similarity information corrected by the case-like similarity government. 4. The method according to any one of claims 1 to 3,
Attribute similarity miscalculation of the similarity evaluation module,
A reference setting section for optimizing the attribute similarity criterion (MCAS) of design summary information by executing a genetic algorithm based on the attribute similarity criterion (MCAS) of design summary information for various information on the existing project stored in the case database;
A calculation unit for storing the function of Equation 1 and receiving and calculating a property similarity criterion (MCAS) set through the reference setting unit and design outline information of a new project and a ready-made project;
Construction cost prediction system for a building, characterized in that consisting of. In claim 4,
The standard calculation section,
A construction cost prediction system for a building, characterized by optimizing the attribute similarity criterion (MCAS) combination of design summary information by executing a genetic algorithm based on the MCAS combination. 4. The method according to any one of claims 1 to 3,
The case database,
The construction cost information is configured to include the information broken down by direct construction costs by type,
The case database and the project input module,
It is an optional influencer that selectively affects the construction cost estimation by adding essential design factors or special design conditions or differences in construction type. On the other hand, the impact factors are configured to be set separately according to the type of work,
The similarity evaluation module,
The attribute weight calculation government is configured to calculate the attribute weight value for each sector by executing genetic algorithms for each sector, targeting the design overview information of the attribute that affects the direct construction cost for each sector.
The similar case generation module,
It is configured to generate similarity integrated information of ready-made project including direct construction cost information and case similarity information per floor area of each ready-made project.
The construction cost calculation module,
The construction cost prediction system of a building, characterized in that it is configured to calculate the construction cost of a new project as the direct construction cost by type by receiving the direct construction cost information per floor area of each existing project. In claim 6,
The case database,
The construction cost information is configured to further include indirect construction cost information provided as a rate for the sum of the direct construction costs,
The similar case generation module,
It is configured to generate the similarity integrated information of the ready-made project including the indirect construction cost information of the ready-made project,
The construction cost calculation module, the construction cost prediction system of a building, characterized in that configured to calculate the indirect construction cost of the new project by receiving the indirect construction cost information of the existing project. As a method for estimating and predicting construction costs at the stage of planning a new project for a building,
Step S11 of constructing a case database in which design summary information and construction cost information for each ready-made project of the completed building are stored as numerical information;
A step S12 of optimizing the attribute weight combination for each attribute of the design summary information by executing a genetic algorithm for a generation of attribute weight combinations for each attribute of the design outline information for the various information on the existing project stored in the case database;
Step S21 of receiving design summary information for each property of a new project of a building;
Step S31 of calculating the property similarity diagram of the ready-made project for the new project by the properties of the design-information information by linking the design overview information of the new project received and the design overview information of the ready-made project stored in the case database with [Equation 1]. ;
A step S32 for calculating a case similarity diagram of a ready-made project by using the property weight value optimized in the step S12 and the property similarity degree calculated in the step S31;
Step S41 of integrating various information on the ready-made project stored in the case database with case similarity information calculated in step S32 to generate similarity integrated information of the ready-made project including construction cost information and case-likeness information of the ready-made project;
A step S42 of calculating a construction cost of a new project using construction cost information per mileage of any of the ready-made projects among the similarity integrated information of the ready-made project generated in step S41;
Construction cost prediction method for a building, characterized in that comprises a.
[Formula 1]

here, f _AS (x) = property similarity estimation function
AV _{Test _ Case} = Design Outline by Property of New Project
AV _{Retrieved _ Case} = Design outline information by property of established project
MCAS = set attribute similarity criterion (0 ~ 1)
[Formula 2]

Where f _CS (x) = case similarity estimation function)
f _ASi = Attribute Similarity _{Diagram by} Design Outline Information
f _AWi = Attribute Weight by Design Outline 9. The method of claim 8,
In step S32,
Step S32a, after calculating case similarity of the ready-made project, extracting various kinds of information of the ready-made project belonging to a higher level;
A second step of optimizing the attribute weight combination for each attribute of the design summary information by executing a genetic algorithm for the generation of attribute weight combinations for each attribute of the design outline information for the various information on the ready-made project extracted in step S32b;
And a step A3 of correcting the case similarity of the established project by the above [Equation 2] using the attribute weight value optimized in the step S32c and the attribute similarity calculated in the step S31.
In step S41,
The construction cost prediction method of a building comprising the similarity integrated information of the ready-made project as various information about the ready-made project extracted in step S32a and the case similarity information corrected in step S32c. The method of claim 8 or 9,
In step S12,
Step S12a optimizing the attribute similarity criterion (MCAS) of the design summary information by executing a genetic algorithm based on the attribute similarity criterion (MCAS) of the design overview information for various information on the existing project stored in the case database; It is made to include more
In step S31,
A method of predicting the construction cost of a building, characterized in that the property similarity is calculated according to [Equation 1] using the optimized property similarity criterion (MCAS) in step S12a.

Images