摘要
特征选择是一种处理维数约简的有效方法。以条件熵为特征子集评价条件,采用随机搜索和启发式搜索相结合的搜索策略,设计了一种新的特征选择方法。该方法不仅能够求得经典启发式特征选择方法的选到特征子集,还可以得到一些与其不同的满足条件特征子集,同时在多数情况下可以减少时间消耗。实验研究表明了提出的算法的有效性。
Feature selection is an important method to reduce dimension of data sets. In this paper, a new feature selection approach is proposed, in which the conditional entropy is employed to evaluate the validation of feature subset, and a search strategy merging random strategy and heuristic strategy is used to decrease the search space. The result obtained by using original algorithm is contented in the one by operating the proposed algorithm, and the consuming time, in most instances, is less than the original one. The experiment indicates that the algorithm is effectiveand efficient.
出处
《太原科技大学学报》
2010年第5期413-416,共4页
Journal of Taiyuan University of Science and Technology
关键词
特征选择
条件熵
随机搜索
启发式
feature selection, conditional entropy, random searching, heuristic