基于模拟退火算法和最近邻分类器识别率的特征选择方法

Feature Selection Based on Simulated Annealing Algorithm and the Recognition Rate of the Nearest Neighbor Classifier

为了提高分类器的识别率,提出一种基于模拟退火算法和最近邻类器识别率的特征选择方法。该方法采用特征度量中的准确性度量(分类器识别率),能直接评价特征组合的优劣,并结合改进的模拟退火算法使用,能确保收敛到近似全局最优解。仿真试验表明,该方法所求解的质量优于传统特征选择方法,并且有很强的噪声容忍性,有一定的工程实用价值。

In order to enhance the recognition ratio of the classifier, a new feature selection method is proposed based on the simulated annealing algorithm and the nearest neighbor classifier. The accuracy metric incombination with a modifeid simulated annealing algorithm can insure the convergence on the approximate global optimal solution. Simulation results show that the method is superior to other traditional methods.