Instance selection for time series classification based on immune binary particle swarm optimization

Abstract

We propose a new immune binary particle swarm optimization algorithm (IBPSO) to solve the problem of instance selection for time series classification, whose objective is to find out the smallest instance combination with maximal classification accuracy. The proposed IBPSO is based on the basic binary particle swarm optimization (BPSO) algorithm proposed by Kennedy and Eberhart. Its immune mechanism includes vaccination and immune selection. Vaccination employs the hubness score of time series and the particles’ inertance as heuristic information to direct the search process. Immune selection procedure always discards the particle with the worst fitness in the current swarm for preventing the degradation of the swarm. Experimental results on small and medium datasets show that IBPSO outperforms BPSO and deterministic INSIGHT in terms of storage requirement and classification accuracy, and presents better robustness to noise than BPSO. In addition, experimental results on larger datasets indicate that IBPSO has better scalability than BPSO.