基于互信息和遗传算法的两阶段特征选择方法

为了在特征选择过程中得到较优的特征子集,结合标准化互信息和遗传算法提出了一种新的两阶段特征选择方法。该方法首先采用标准化的互信息对特征进行排序,然后用排序在前的特征初始化第二阶段遗传算法的部分种群,使得遗传算法的初始种群中含有较好的搜索起点,从而遗传算法只需较少的进化代数就可搜寻到较优的特征子集。实验显示,所提出的特征选择方法在特征约简和分类等方面具有较好的效果。

Two-stage feature selection algorithm based onmutual information and genetic algorithm

To get better feature subset in the feature selection process, this paper proposed a new two-stage feature selection algorithm based on normalized mutual information and genetic algorithm. First it ranked features by normalized mutual information. Then to provide the genetic algorithm with better starting point it used the front ranking features to initialize the population, thus got better feature subset after only a few evolution times. The test results on benchmark datasets show the effectiveness of the algorithm, in terms of dimensionality reduction and classification performance.