An Adaptive Particle Swarm Optimization Algorithm for Enhancing AI-Driven Time Series Forecasting and Data Analytics

Abstract

Particle Swarm Optimization (PSO) is a metaheuristic optimization method that has attracted a lot of interest as an optimisation method that is simple, flexible, and has high global search capacity to solve complex engineering problems. Nonetheless, the traditional PSO algorithms tend to be prematurely convergent and limited in their exploration capacity when used on high-dimensional and nonlinear optimization problems. This paper presents a dynamic Adaptive Particle Swarm Optimization Algorithm (APSO) that varies several important control parameters, such as inertia weight and acceleration coefficients, based on the search progress and diversity of the population. The adaptive mechanism enhances the trade off between exploration and exploitation, and the particles can escape local optima and converge more quickly.