A project risk forecast model based on support vector machine

Abstract

A Project risk forecast model was investigated using least square support vector machine(LS-SVM) method. Risk estimation data of experts was acted as eigenvector of learning samples to train the constructed LS-SVM regression model for realizing mapping relationship between the risk and the characteristic. The test samples were used to compare between the constructed LS-SVM model and BP neural network. The result showed that LS-SVM model has high prediction accuracy and strong generalization ability. So it is suitable for the forecast of large scale project risk.