MODELLING CLIENT SATISFACTION LEVELS: A COMPARISON OF MULTIPLE REGRESSION AND ARTIFICIAL NEURAL NETWORK TECHNIQUES

The performance of contractors is known to be a key determinant of client satisfaction. Here, clients ' satisfaction is defined in several dimensions identified using factor analysis techniques. Based on clients ’ assessment of contractor performance, a number of satisfaction models are presented, developed using multiple regression (MR) and artificial neural network (ANN) techniques. The MR models identified that various attributes of the contractor, project and client were found to significantly influence satisfaction levels. Results of the ANN modelling were similar, however the importance of independent variables was found to be different. The models demonstrate accurate and reliable predictive power as confirmed by validation tests. While the MR models tend to be more accurate for specific dimensions of client satisfaction, the ANN models were found to be superior for models of average satisfaction and overall satisfaction. The MR models suggest that contractors have more effect on client satisfaction than the ANN models. Contractors could use the models to help improve their performance leading to more satisfied clients. This will also promote the development of harmonious working relationships within the construction project coalition