A multi-information fusion “triple variables with iteration” inertia weight PSO algorithm and its application

Abstract

Particle swarm optimization (PSO) has many advantages such as fewer parameters, faster convergence and easy implementation; however, it is also prone to fall into local optimum. Because inertia weight parameters can increase the diversity of particles and effectively overcome this problem, a large number of studies have done on inertia weight strategy since it was put forward, and many different improvement strategies have been put forward, but these improvement strategies still do not make full use of the multi-dimensional information of particles, resulting in limited improvement of PSO performance. In this study, on the one hand, based on the summary of current inertial weight strategies, aiming at the problems existing in the classification of inertial weight strategies, the inertial weight strategies are reclassified. According to the new classification methods, the inertial weight strategies are divided into four categories, including “no variable with iteration”, “single variable with iteration”, “double variables with iteration” and “triple variables with iteration” inertia weight. On the other hand, in view of the shortcomings of the existing inertial weight improvement strategies, this study proposes a multi-information fusion “triple variables with iteration” inertia weight PSO algorithm (MFTIWPSO), which combines the multi-dimensional information of particle’s time (or iteration), particle and dimension (or space). In order to test the optimization performance of our proposed MFTIWPSO, the benchmark functions are used to test the optimization performance, and then the algorithm is used to optimize the parameters of machine learning classifier and classify the biological datasets. The results of two tests show that the proposed MFTIWPSO algorithm has better optimization performance than other optimization algorithms.