摘要
针对特征选择中降维效果与分类精度间的矛盾,通过分析传统的特征选择方法中的优点和不足,结合佳点集遗传算法的思想和K最近邻简单有效的分类特性,提出了基于佳点集遗传算法的特征选择方法。该算法对特征子集采用佳点集遗传算法进行随机搜索,并采用K近邻的分类错误率作为评价指标,淘汰不好的特征子集,保存较优的特征子集。通过实验比较看出,该算法可以有效地找出具有较高分类精度的特征子集,降维效果良好,具有较好的特征子集选择能力。
To address the contradiction between the dimension reduction for feature selection and the precision of classification, by analyzing the strengths and weaknesses of the traditional feature selection method, combines the idea of good point-set genetic algorithm and the simple and effective features of K nearest neighbor classification ,presents a new feature selection method based on good point set genetic algorithms. Through a random search of the feature subset with the good point-set genetic algorithm, and using K nearest neighbor classification error rate as the evaluation index, eliminate the bad feature subset, save the optimum feature subset. It can be seen through the comparison experiments that the algorithm can effectively find out those feature subset which has high classification accuracy, and the effect of dimension reduction is good, these show that the algorithm has the better ability to select feature subset.
出处
《计算机技术与发展》
2011年第1期50-52,57,共4页
Computer Technology and Development
基金
安徽省高等学校省级自然科学基金(KJ2008B092)
关键词
K最近邻算法
特征选择
佳点集遗传算法
K-nearest neighbor algorithm
feature selection
good point-set genetic algolithm