Abstract: To address the significant impact of features selection and parameter optimization on classification accuracy in support vector machine (SVM), a feature selection algorithm based on the improved sootytern algorithm to optimize SVM synchronously is proposed. Firstly, the Tent chaotic map is used to initialize the sooty tern population to increase the diversity of the population. On this basis, cosine adaptation and simulated annealing algorithm are introduced to avoid the defect of the sooty tern algorithm falling into local optimality, enhance the global search ability of the algorithm, and improve the convergence accuracy. Secondly, the improved algorithm is combined with feature selection and SVM, and the parameters of binary feature selection and SVM are optimized simultaneously. Finally, feature selection simulation comparison experiments are carried out on 10 standard datasets. Compared to the original and other optimization algorithms, the proposed algorithm can effectively reduce the data dimension and improve the classification accuracy.

Key words: sooty tern optimization algorithm, cosine adaptation, simulated annealing algorithm, support vector machine, feature selection