分类问题中的特征选取  

Feature Selection in Classification Problem

作　　者：高文卿[1] 赵永升[2] 

机构地区：[1]山东商务职业学院信息与艺术设计系,山东烟台264670 [2]鲁东大学信息与电气工程学院,山东烟台264039

出　　处：《鲁东大学学报（自然科学版）》2014年第3期219-222,共4页Journal of Ludong University:Natural Science Edition

摘　　要：针对分类问题提出了一种特征选择的新算法.算法在初始化时首先任意产生几个特征集,然后被迭代多次执行,在每一次迭代过程中,用性能评估函数对所有的特征集进行评价,按照评价结果选取当前性能最优的特征集,其它所有的特征集朝着类似当前最优特征集的方向发生变化,直至完成预定的次数为止.According to the classification problem, a new algorithm of feature selection is proposed. First of all, several feature sets randomly are produced when the algorithm is initialized, then it will be iteratived times. All the feature sets are evaluated by using the performance evaluation function in each generation. The optimal current feature set is selected according to the evaluation results, and all the other feature sets change towards the similar direction of the optimal current feature set until the algorithm is performed in pre-defined times.

关 键 词：特征选取 分类问题 迭代 评估函数 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]