深度优先搜索的支持向量机参数优化算法

Study on Parameters Optimization of Support Vector Machines Based on DFS

研究支持向量机参数优化问题,由于算法要求准确选择SVM参数,支持向量机在处理大样本数据集时和最优模型参数确定时,消耗的时间长、占有内存大,易获得局部最优解的难题。为了解决支持向量机存在的不足,采用深度优先搜索算法对其参数优化机机制进行改进。将向量机参数优化视成一个组合优化问题,将支持向量机模型的分类误差作为优化目标函数,采用深度优先算法对其进行求解,最后将模型应用于3个标准分类数据集。仿真结果表明,优化参数后的支持向量机加快模型的训练速度度,提高了分类的准确率,很好的解决了支持向量机参数优化难题。

Study on the problems of support vector machines parameters optimization.The prediction precision of Support vector machines model and generalization ability depend on its parameters reasonable choice.The problems of time-consuming and easy falling into the local optimal value exist in traditional support vector machine parameters optimization algorithm,and support vector machine prediction precision is low.In order to solve the problems,the paper puts forward a method based on depth first search of SVM parameters optimization method（DFS-SVM）.DFS-SVM takes SVM parameters optimization as a combinatorial optimization problem and the RMSE as optimization goal,uses depth first search to select SVM parameters,and tests DFS-SVM performaces through three standard data set.Simulation experiment results show that the DFS-SVM prediction accuracy is improved and the training time is shorten greatly.It provides a new effective solution for SVM parameters optimization problem.