Cost Controller for Construction Projects Based on Fuzzy Least Squares Support Vector Machines

In this paper, we propose a hybrid cost controller to deal with complicated attributes and high controlling risk of construction cost based on fuzzy least squares support vector machines. In this controller, fuzzy membership and least squares support vector machines are combined together. Considering the specificity of project sample data, complex fuzzy membership function is employed to improve their representation abilities, in which cost attributes are reconstructed. Then, east squares support vector machines is trained by these obtained optimum samples and its prediction data are regarded as the comparative data sources for cost controlling, which can make early risk warning by the appropriate threshold value designed in the model. As a result, LS-FSVMs project cost controller is built. The proposed model not only makes a good use of the fuzzy theory ability to deal with uncertain things, but also takes advantage of LS-SVMs ability to solve the problem with small samples and nonlinear regression. Furthermore, the proposed approach is shown more accurate for prediction in the case of real-word application.