Research On Quality Cost Forecasting Of Construction Projections Based On SVR

As one of the six links of the quality cost management of the project, the quality cost forecasting of construction projections is the basis of the effective implementation of the other links and it also is the basis of cost of quality managers in decision-making options. The high-activity predicting of the quality cost is necessary to the practice and theory of project quality cost management. According to the present research of the building quality cost, in the thesis, found some of the problems existing on quality cost forecast of the building. Combination unascertained measure and support vector machine (SVM) theory, put forward a new model suitable for the quality cost predicting of construction projections.First, for the lack of traditional forecasting methods combined with the features and advantages of support vector machines, it conducted a comprehensive study to learning rules and the training process of the SVM. The support vector machine can solve the global optimal solution in the small sample data, but ingenious solution to the problem of dimension that it is not related between the complexity of the algorithm and sample dimension. Meanwhile, the SVM model overcome some disadvantages such as traditional neural network algorithm which is difficult to determine the hidden layer, and is easy to fall into local minima as well as lower prediction accuracy.Secondly, according to internal and external affecting factors the quality cost of construction, to establish a more reasonable indicators system of the quality cost predicting of construction projections, and made a prediction model of the quality costs based on support vector regression machine (SVR).In view of this method for determining the SVR model parameters and kernel function is still no uniform theoretical principles. Conducting comparative analysis of predicted the results on different parameters of optimization methods and different kernel functions, obtained the best SVR model which is suitable the quality of construction cost projections.Finally, explored the steps of the SVR model in MATLAB software, and the model is applied to the instance, proved the feasibility and higher accuracy of the model.