Feature Selection Method Based on Hybrid Multi-Strategy Pelican Optimization Algorithm

Pelican Optimization Algorithm is a new stochastic natural heuristic optimization algorithm with good global development capability. However, the pelican algorithm has the problem of low convergence accuracy and easy to falls into local optimization in feature selection. In this paper, a hybrid multi-strategy pelican optimization algorithm (HMS-POA) is proposed. First, in the initialization phase, dynamic reverse learning and tent chaotic mapping are introduced to increase population diversity and avoid falling into local optimum. Secondly, in the first stage, a random movement strategy is selected, and in the second stage, the harris hawk algorithm is introduced. Compared with other optimization algorithms on 13 benchmark functions, the results show that HMS-POA performs better. The data classification research is conducted through 14 groups of classical UCI datasets. The experimental results demonstrate that the algorithm performs better and more competitively in delivering the optimum solution for optimization problems. It can analyze data more precisely to prevent redundant feature interference.