A Dynamic Adaptive Particle Swarm Optimization for Knapsack Problem

Concepts such as equivalent value transformation, reverse value transformation and transform sequence were defined according to characteristics of 0-1 knapsack problem, then a special particle swarm optimization was proposed to solve knapsack problem. The random inertia weight was introduced which made an ideal balance between the capability of global exploration and the capability of local exploitation, and dynamic adaptive mutation was adopted to reinitialize part of swarm when the algorithm slump into local best fitness value which avoided trapping to premature convergence. The experimental results show that the algorithm can evidently alleviate the undulate phenomenon in the evolutionary process, therefore improve the stability of particle swarm optimization, and increase the convergent velocity and precision.