一种支持向量机更新模型的参数选择方法

A Method for Parameter Selection of Support Vector Machine Updated Model

支持向量机性能主要受模型参数的影响,而支持向量机更新模型的参数选择尚无专门的方法。将量子遗传算法用于模型参数选择并进行改进,用原始模型最优参数设置初始量子模板的生成规则,确定搜索方向;用拟合误差设置量子旋转门的调整策略,缩小解空间取值范围。通过仿真验证了所提方法的有效性。该方法能有效搜索到最优参数,与基本遗传算法相比,其解的精度在搜索过程的初期较高,搜索代数大大降低,能有效降低运算量。

The model parameters have great effect on the performance of Support Vector Machine（SVM）,but there＇s no special method for model parameter selection of SVM.An improved Quantum Genetic Algorithm（QGA） was proposed,the optimal parameters of the original model were used to set up the production rules of initial quantum template,thus the search direction was defined.The fitting error was introduced to design the adjust step of quantum rotation door,thus solution space was narrowed.Simulation proved the effectiveness of the proposed method.Compared with genetic algorithm,the proposed method can search optimal parameters correctly,reduce search generations and computation effectively,and has high accuracy at early stage.