In the construction industry, decision making in the early phase of a project is a very important factor in the entire project’s success. However, the information available that can be used in the early stage is limited, making it difficult to accur...
In the construction industry, decision making in the early phase of a project is a very important factor in the entire project’s success. However, the information available that can be used in the early stage is limited, making it difficult to accurately predict the budget needed for construction. For this reason, many researchers utilize various methodologies in estimating the construction budget at the early stage of construction.
Recently, a construction budget prediction model using ‘Case-based Reasoning (CBR)’ method has elicited much attention, and there have been continuous studies on improving the accuracy of CBR-based prediction models. In such studies, the prediction was made with the information fixed in the early stage of construction. However, due to the characteristics of the early stage of a project, there are considerable uncertainties. Also, since design alterations are common, there are limitations for such budget prediction models when it comes to decision-making.
This research utilized Monte Carlo Simulation (MCS) to consider the probability of available information with uncertainties in the early stage of a project. A probabilistic prediction model for construction budget using Monte Carlo Simulation and linking the methodology to Case-based Reasoning is proposed.
The main points in this research are summarized below.
First, this paper analyzed earlier researches on existing methodologies to predict construction budget, with earlier researches using Monte Carlo Simulation. The methodology utilized in this research is explained in detail.
Second, through situation analysis of existing CBR-based construction budget prediction models, the research identified problems and limitations of existing models, arguing the need for improvements.
Third, the paper suggested a method to apply MCS for CBR-based construction budget prediction models and developed a probabilistic, hybrid, construction budget prediction model. The paper conducted regression analysis based on the database built to attribute weighting of influence factors to be used in the retrieval stage in CBR. A hypothesis was created on the distribution types for each influence factor the findings of earlier researches. Through distribution types and fluctuation range for each influence factor, a random number combination of construction budget was created and used as basis for a similar case inquiry. The process of random number combination creation and inquiry process of similar cases was repeated at a regular basis. Based on the accumulated inquiry result, a probabilistic construction budget prediction result was extracted.
Fourth, the research suggests average, mode, median value and budget range from the probabilistic construction budget prediction result for a project requiring budget prediction. Moreover, a probability of successful construction with a designated budget value could be given. Different budget values could also be given based on the risk.
Fifth, case collection of apartment houses was done to check the effectiveness and validity of the probabilistic construction budget prediction hybrid model. Findings were verified using the average value taken by considering uncertainties, reflecting all the cases in the prediction result. A comparative analysis for verification was conducted for cases including ① error rate (A) of construction budget prediction using CBR method and fixed influence factors; ② error rate (B) of prediction using regression analysis with fixed influence factor values; ③ error rate (C) of the result from using probabilistic values of influence factors; and ➃ error rate (D) for application of probabilistic method used for only a part of the influence factors.
This research randomly selected 500 cases from 510 construction cases. The 10 leftover cases were used in the verification process. The average error rate on verified cases are as follows: error rate of method (A) that utilizes CBR and fixed influence factors is 8.9%; the construction budget error rate using regression analysis and fixed influence factors (B) is 7.0%; the result of probabilistic construction budget prediction model (C) showed a 3.5% error rate; and the error rate for application of probabilistic method used for only a part of influence factors (D) is 5.2%. Two cases using the probabilistic construction budget prediction model showed a relatively low error rate, therefore indicating high prediction performance.
Unlike existing construction budget prediction models that require fixed values of influence factors, the hybrid construction budget prediction model suggested in this research can use the probability distribution and fluctuation range of influence factors to predict construction budget with higher accuracy. This model is expected to be useful in the early phase of projects when available information is limited and more uncertainties are present. Decision-making of ordering bodies is expected to benefit from the advantages of this method.
A limitation of this research was that it was conducted on the retrieval stage of CBR, and correction was not done for inquired cases. A correction process would be needed in future researches, and detailed validity analysis of average values used for prediction result values in this research is also needed. Moreover, database building and additional verification for cases other than apartment houses facilities should continue.