Function mining based on gene Expression Programming and Particle Swarm Optimization

Gene expression programming (GEP) is a powerful tool widely used in function mining. However, it is difficult for GEP to generate appropriate numeric constants for function mining. In this paper, a novel approach of creating numeric constants, GEPPSO, was proposed, which embedded particle swarm optimization (PSO) into GEP. In the approach, the evolutionary process was divided into 2 phases: in the first phase, GEP focused on optimizing the structure of function expression, and in the second one, PSO focused on optimizing the constant parameters. The experimental results on function mining problems show that the performance of GEPPSO is better than that of the existing GEP random numerical constants algorithm (GEP-RNC).