A method for feature selection based on the optimal hyperplane of SVM and independent analysis
Lin-Fang Hu, Wei Yu Gong, Li-Xiao Qi, Ping Q. Wang · 2013
Feature selection is an important topic in machine learning. In order to evaluate the candidate features, a strategy based on the constituent principle of the SVM optimal hyperplane is established in this paper. Then, by considering different feature combinations, a better feature subset can be obtained. The method is used to recognize the monomers in weather forecast, and experimental results demonstrate its effectiveness in enhancing the classification performance.