Mutation grey wolf elite PSO balanced XGBoost for radar emitter individual identification based on measured signals

Abstract

• Radar individual identification depends on signal subtle features. • Number of radar samples measured from real environment is imbalanced. • Parameters of XGBoost have great influence on the identification results . • PSO sometimes falls into local optimal solution. Radar emitter individual identification plays an increasingly important role in electronic support measures (ESM) system. To cope with the problems of low accuracy and poor stability of radar emitter individual identification, a novel method, named MGWEPSO-BXGBoost (Mutation Grey Wolf Elite Particle Swarm Optimization Balanced eXtreme Gradient Boosting), is proposed. In consideration of the fact that the number of radar signals measured from the real environment is usually imbalanced, a novel balance mechanism is designed for XGBoost. MGWEPSO is further proposed to simultaneously optimize the prime parameters, whose values and combinations have a great influence on the identification results, of BXGBoost to improve the identification accuracy. To overcome the local optimal solution problem, leadership mechanism in wolves, elite rule and the idea of mutation are adopted, which is also conducive to improving the ability to find the global optimal solution. Furthermore, experiments based on signals measured in the real environment are carried out to demonstrate the effectiveness of the proposed method. The results verify that MGWEPSO-BXGBoost has high accuracy and strong stability even when the sample size of each individual is limited and imbalanced.