Abstract:Support Vector Machine (SVM) classifier algorithm,which can map the original features by the kernel function to solve the linear inseparability problem in the original space,has excellent performance in high-dimensional data classification.However,due to the different data characteristics and dimensions,it is generally necessary to manually test the parameters of SVM with low efficiency and increased workload.To address this problem,this study proposes an optimization method for SVM classifier based on Scikit-learning.The optimal parameter range is searched by grid search method,and the classification results of different parameters and different kernel functions are visualized in Scikit-Learn of Python machine learning library.The optimal parameter range is displayed on the grid to find the parameter distribution with high accuracy.The parameters are solved more accurately by means of automatic iteration,and the corresponding values are set to be brought into the iteration calculation.Moreover,the neighborhood search of the optimal parameters is determined to choose the optimal parameters value directly,in order to avoid falling into over-fitting.Experimental results show that the time of manual parameter adjustment is reduced greatly and the accuracy of the optimal parameters obtained here is improved.