Research on Prediction Model Based on Hybrid Algorithm of PSO-SVM Kernel Function

To solve the problem of low prediction accuracy of SVM algorithm, this paper proposes a prediction model research method based on PSO-SVM kernel function hybrid algorithm, designs global and local kernels hybrid kernels, and uses PSO algorithm to optimize SVM algorithm. The hybrid kernel function is used to improve the prediction accuracy. Using the RSS simulation data as the training sample, through comparing the SVM algorithm, PSO-SVM algorithm and PSO-SVM kernel function hybrid algorithm, the simulation results show that the prediction accuracy based on PSO-SVM kernel function hybrid algorithm is the highest rMSE = 29.7408, followed by PSO-SVM algorithm rMSE = 30.7976, and finally SVM algorithm, rMSE = 39.2858.