Research on Intelligent Prediction Method of Engineering Cost System Based on SVM Algorithm

Support Vector Machine (SVM) is a new machine learning technology developed based on statistical learning theory. It has a solid theoretical foundation, superior learning ability and good generalization ability. It is based on the principle of minimizing the structural risk and does not require much sample information. Therefore, the best compromise between the complexity and learning ability of the model can be found by using a small amount of sample information. Training SVM is equivalent to solving a quadratic programming problem with linear constraints, and there is a unique solution. In this paper, the basic principle of SVM is analyzed. SVM has the remarkable characteristics of nonlinear fitting, strong generalization ability and fast training convergence speed. Aiming at the nonlinear relationship between engineering cost system and various influencing factors, this paper puts forward a new short-term forecasting method of engineering cost system based on SVM by using its superior nonlinear learning and forecasting performance, so as to improve the forecasting accuracy and timeliness. This paper establishes a forecasting model of engineering cost system based on SVM, and compares it with neural network method. The results show that the forecasting accuracy and speed based on SVM are better than neural network method. In this paper, a prediction method combining F -score feature selection with SVM is proposed. This method uses F -score feature selection to reduce the input feature dimension and then sends it to SVM for modeling, which not only combines the feature selection ability of F -score but also makes use of SVM's good nonlinear function approximation ability, thus improving the accuracy and generalization ability of prediction model. It is of practical significance to improve the intelligent prediction level of China's engineering cost system.