Improving accuracy of early stage cost estimation by revising categorical variables in a case-based reasoning model

For the overall success of a construction project, it is very important to accurately estimate the construction cost from the early stage. However, the limited information available in the early stage makes the cost estimation challenging. Recently, there has been an increase in the use of case-based reasoning (CBR) to estimate construction cost in the early stage. Based on the hypothesis that similar problems have similar solutions, CBR searches for the most similar cases (e.g., previous projects) for a given problem (e.g., a new project). However, because no previous project is exactly the same as a new project, the solutions applied to the past project may not work for the new project, especially when there are no sufficient cases stored in the case base. To overcome this limitation, some studies have highlighted the importance of revision algorithms to account for the deviation of the new project and the identified similar past projects. These studies, however, were limited in considering the deviation of numerical variables while the majority of variables available in the early stage is more categorical (e.g., structural system or underground condition) than numerical (e.g., gross floor area). Based on this recognition, this paper presents a revision method considering the deviation of categorical and numerical variables using regression analysis. Application to multihousing projects confirmed that the proposed CBR model can increase the accuracy of construction cost estimation. This paper is of relevance to researchers in terms of providing a theoretical basis to incorporate both numerical and categorical variables in revising the CBR model. This paper is also of value to practitioners with regard to providing an accurate cost estimation tool in the early stage of construction projects.