关键词: 支持向量机, 模型选择, 网格模式搜索

Abstract: For fixed functional form of the kernel，model selection amounts to tuning kernel parameters and the slack penalty coefficient C.Based on an analysis of the grid algorithm and pattern algorithm，this paper proposes a grid-pattern search algorithm，which combines grid search and pattern search.The main procedure of the proposed method include a fast search in the global domain with grid algorithm，then after obtaining the least interval containing the optimal solution，a pattern algorithm is employed to get the optimal solution in the interval.Experimental results indicate that this method has the advantage of high accuracy and speed when training SVM.

Key words: support vector machine, model selection, grid pattern search