Forecasting contractor's deviation from the client objectives in prequalification model using support vector regression

Contractor prequalification (CP) is a very complex decision-making process with nonlinearity, uncertainty and imprecision in inputs containing both subjective and objective data. The failure to perform CP can lead to large losses, delays or severe loss of project quality. Although the most reliable approach identified in the literature is currently artificial neural network (ANN), it has weaknesses that negatively affect CP. In this study, a new approach called support vector machines (SVM) has been used to forecast a contractor's deviation from a client's objectives. In order to test the model, CP for 250 virtual contractors was solved. The proposed model had a great generalization in linear, nonlinear, noisy and inductive environments. The Results showed that SVM could reliably perform even with a small amount of training data. Also when compared to ANN, SVM showed an overall better performance.