Project cost forecasting is a key point in the research on project management, in view of support vector machine parameter optimization problem in project cost forecasting, a new project cost forecasting model (IPSO-SVM) is proposed, which is based on the improved particle swarm optimizing supporting vector machine. Firstly, project cost data is collected and processed, and then support vector machine is used to learn for training samples in which improved particle swarm algorithm is used to optimize kernel function parameters of support vector machine, At last, the simulation experiment is used to test the performance of project cost forecasting by using Matlab 2012. The experimental results show that IPSO-SVM can effectively improve the forecasting accuracy of project cost, and the forecasting results have some practical application values.