Modelling construction labour productivity from labour's characteristics

Labour is a fundamental input to any construction project to achieve the highest level of productivity. Productivity remains as one of the most important ways to measure the overall performance of construction project. Construction productivity is directly related to labour and thus, it is mainly dependent on human effort and performance. Improvement of Construction Labour Productivity (CLP) can directly help to improve the performance of construction companies, become more competitive, besides contributes to national economy. The aim of the research is to develop and introduce a new framework for systematic assessment of the factors influencing construction labour productivity and use the collected data to create models by applying state-of-art techniques and comparing the accuracies in predicting the labour productivity in construction. The scope of the study was limited to Malaysia only. A thorough literature survey was conducted to list the factors related to CLP with different studies throughout the globe. The factors were filtered using two-stage procedures - first the factors were shortlisted based on the relevance of labour and then a survey was conducted among project managers to rank the factors based on the importance of Malaysian context using a 3-point Likert scale on each factor. The ranks of the factors were analysed using statistical tools. The top class factors were identified using Jenks Optimization Techniques. The classified CLP factors were used to design a field survey to collect data from construction workers. Five state-of-arts of models were developed to predict the CLP from the factors including three data mining models, one conventional model and one multi-criteria model. Salary of labour was considered as a proxy to the productivity to develop the models. The performance of the models were assessed using five categorical indices. The results of literature review revealed that a total of 112 factors related to productivity in construction industry have been identified throughout the globe. Ten factors were identified through the analysis of preliminary survey data using different methods. Among them, seven factors were found common for all the methods which were identified as the important CLP factors for Malaysian construction industry. The factors are (1) Lack of Work Experience (2) Job Category (3) Education/Training (4) Nationality (5) Worker Skills (6) Age and (7) Marital Status. The relative performance of different models was compared to identify the best model in term of the rate of accuracy in prediction of labour productivity. Data mining models were found to perform better compared to other models. The Percentage of Correct (PC) for data mining models were found in the range of 0.735-0.835, Probability of Detection (POD) between 0.741 and 0.911, Heidke Skill Score (HSS) between 0.792 and 0.802 and Peirce Skill Score (PSS) in the range of 0.792 to 0.799, while the False Alarm Ratio (FAR) were found in the range of 0.102 to 0.279. The values were found better than that obtained using Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) (PC=0.739, POD=0.740, HSS=0.794, PSS=0.725 and FAR=0.256) and much better than that obtained using Linear Regression (LR) (PC=0.577, POD=0.618, HSS=0.533, PSS=0.498 and FAR=0.533). Among the data mining models, Support Vector Machine (SVM) was found to provide the best results in term of all statistical metrics used. The POD for SVM was found above 90% in predicting different categories of productivity. The method discussed in this research can serve as a newly developed framework to predict the level of construction labour productivity for project