Application Research Of Support Vector Machine Based On Kernel Parameter Optimization In Multi-Classification Problems

The ability of SVM to solve the classification problem effectively attracts the attention of many researchers.The application of the support vector classifier model has become a new research direction and focus area.Due to the classification problem,the sample data set is linearly inseparable.The characteristics of multi-classification and the problem of linear inseparability of sample space are solved by kernel technique,which makes the support vector model have great difficulty in parameter selection.The choice of kernel function and corresponding parameters determines the classification performance of support vector machine.as follows:Firstly,on the basis of the development process of support vector classifier and the research at home and abroad,the data structure of the classification problem of different support vector classifier models is analyzed.The evaluation index of multi-classification model is studied.Secondly,based on the analysis and discussion of grid search method,particle swarm optimization algorithm and genetic optimization algorithm,an improved particle swarm optimization algorithm based on grid search method is proposed.The optimized support vector machine model is applied to In the seed classification problem,the seed dataset after data preprocessing is compared by experiments.The polynomial kernel support vector machine classification model has advantages over other algorithms in terms of overall classification accuracy and individual classification accuracy.Finally,for the customer classification problem,the box pre-processing method and the most value normalization method are applied in the data pre-processing stage for data pre-processing,and then the principal component analysis is used to reduce the dimensionality of the data.The Gaussian kernel and polynomial are compared through experiments.Kernel support vector machine classification model,as well as grid search method,particle swarm optimization algorithm,genetic optimization algorithm three optimization algorithms.It is concluded that the Gaussian kernel support vector machine model has better customer classification problem based on genetic optimization algorithm.Classification effect.