Feature selection algorithm based on dynamic programming and comentropy
Zhao Fang · Jisuanji gongcheng yu sheji · 2010
To solve complexity problem of analyzing and dealing with small samples and multi-features dataset,improve complication of data mining and analysis,advance the disposal speed of algorithm at data mining,feature selection algorithm based on dynamic programming and comentropy is proposed.This method based on dynamic programming indicates the capability of chose feature-space with the help of the separability criterion based on the concept of comentropy and reduce dimension of feature-space.Finally,the algorithm is verified through using a part of UCI standard database and the result is compared to other well-known algorithms,experiment shows that the proposed algorithm could reduce effectively not only the dimension of small samples and multi-features dataset and but also the computational cost compared to other optimization seeking algorithm.