Unsupervised Clustering Model for Evaluating the Complexity of Building Production: Building Production Impact Model (BPIM) and Building Production Impact Index (BPI)

Project: Research

View graph of relations


Project managers are aware of the importance of evaluating the complexity of building production during the building production planning stage. However, they admit that shortfalls exist in evaluations as only general project information such as contract duration, contract sum and building areas is usually considered, whereas the evaluation should actually be a complex process involving many variables that affect building production. Previous research has tackled the problem from various perspectives, such as project performance, project success, and contractors’ performance, and has advanced the understanding of the respective domains. However, practitioners comment that previous studies cannot establish an overall picture about the characteristics of building projects and the complexity of building production, and the findings are fragmented. This highlights the need to consolidate previous work and develop innovative new approaches in evaluating the complexity of building production. This project describes the formulation of an industry-wide objective tool to explore the relationship between building project information and building production in terms of a Building Production Impact Model (BPIM) using factor analysis and Self-organizing Map (SOM) unsupervised clustering techniques.Previous researchers generally adopted subjective measures in examining various features of building projects. This study attempts to formulate the BPIM using project information extracted from construction drawings to be gathered from government archives and interviews with contractors. Factor analysis will be used to restructure the variables (the project information) into meaningful factors to review the interrelationship between them. The project samples will then be divided into clusters by the SOM using the factor scores computed by factor analysis. The project clusters can describe the characteristics of building projects from a site production perspective for comparison and evaluation. Subsequently, cluster-based Building Production Impact (BPI) scales and a BPI Index will be derived with reference to the project clusters formed and the baseline reference provided by practitioners. The proposed BPI Index will provide objective measurements, criteria, and standards for evaluating and comparing building projects and guiding managerial decisions in resource allocation. Using factor analysis and the SOM unsupervised cluster technique has the benefits of generating objective, unbiased and long-term models that can be maintained and can be updated from time to time to cope with changes in practice and performance of the construction industry.


Project number9041361
Grant typeGRF
Effective start/end date1/01/0919/03/12