Hybrid Particle Swarm Optimization Feature Selection for Crime Classification

In this study, we propose a hybrid crime classification model by combining artificial neural network (ANN), particle swarm optimization (PSO) and grey relation analysis (GRA). The objective of this study is to identify the significant features of the specific crimes and to classify the crimes into three different categories. The PSO as the feature selection method, reduce the dimension of datasets by selecting the most significant features. The reduction of the datasets' dimension may reduce the complexity thus shorten the running time of ANN to classify the crime datasets. The GRA is used to rank the selected features of the specific crimes thus visualize the importance of the selected crime's attribute. The experiment is carried out on the Communities and Crime dataset. The result of PSO feature selection will then compare with the other feature selection methods such as evolutionary algorithm (EA) and genetic algorithm (GA). The classification performance for each feature selection method will be evaluated. From our experiments, we found that PSO select less features compare with EA and GA. The classification performance results show that the combination of PSO with ANN produce less error and shorten the running time compare with the combination of EA with ANN and GA with ANN.