A Novel Relief Feature Selection Algorithm Based on Mean-Variance Model
Yuxuan Sun, Xiaojun Lou, Bisai BAO - · 2011
The RELIEF algorithm is a popular approach for feature weight estimation. Many extensions of the RELIEF algorithm are developed. However, an essential defect in the original RELIEF algorithm has been ignored for years. Because of the randomicity and the uncertainty of the instances used for calculating the feature weight vector in the RELEIF algorithm, the results will fluctuate with the instances, which lead to poor evaluation accuracy. To solve this problem, a novel feature selection algorithm based on Mean-Variance model is proposed. It takes both the mean and the variance of the discrimination among instances into account as the criterion of feature weight estimation, which makes the result more stable and accurate. Based on real seismic signals of ground targets, experiment results indicate that the subsets of feature generated by proposed algorithm have better performance.