| Summary: | To address the problems of low classification accuracy, redundancy of feature subsets, and performance susceptibility to parameters in wrapper-based feature selection in traditional Support Vector Machine (SVM), an improved Slime Mould Algorithm (ISMA) was proposed for simultaneous optimization of SVM parameters and feature selection. Firstly, an improved Slime Mould Algorithm with multi-strategy was proposed, which has higher convergence speed and accuracy than SMA. Based on the golden section coefficient, a new position updating formula was proposed, which accelerates the convergence speed of SMA and improves the local exploitation ability and convergence accuracy of SMA; based on the idea of Fitness-Distance Balance method, an adaptive lens-imaging learning strategy was proposed, which better balances the exploration and exploitation ability of SMA; the vertical crossover was used to expand the search range, thereby reducing the probability of the algorithm falling into the local optimum. Secondly, ISMA is verified on some standard test functions, CEC2017 test set functions and practical engineering optimization problems. The experimental results show that ISMA has higher solution accuracy, better stability and faster convergence speed, and has higher performance in practical engineering optimization problems. Finally, ISMA was applied to the feature selection process of SVM to optimize SVM and binary feature parameters at the same time, and this method is applied to the microarray gene expression classification problem. The simulation results of feature selection on 10 UCI data sets show that this method can achieve higher classification accuracy while effectively reducing the feature dimension, and the classification accuracy on 7 datasets is as high as 90% above, which reached 100% on 2 datasets. In addition, experiments on two cancer datasets show that this method has good application value in cancer diagnosis and classification.
|
|---|