Feature Selection Algorithm based on Double-hook Function and TWSVM

Twin support vector machine (TWSVM) produces a pair of non-parallel hyperplanes in the progress of classification, which makes each feature correspond to two weights. While using the two weight vectors to select features directly is not possible. In this paper, a feature selection algorithm based on double-hook function and twin support vector machine is proposed, namely double-hook TWSVM algorithm. In this algorithm, combined with the characteristics of the double-hook function, the two weight vectors generated in the process of solving the linear TWSVM are merged into one single weight vector, which represents the corresponding feature’s degree of importance, so that this weight vector can be used to select the characteristics. Finally, the algorithm is analyzed by UCI dataset. To further prove the performance of our algorithm in the aspect of feature selection, it is then used to select color features from a real dataset of woodiness board and make classification. The experimental results show that the feature selection algorithm based on the double-hook twin support vector machine can remove the redundant features, reduce the dimension of features, and ensure a high classification accuracy, which means the algorithm has played a crucial role in the selection of features.