An application of strength pareto evolutionary algorithm for feature selection from crime data

Genetic algorithm is a computational technique that helps to find the optimal solution in the process of natural selection and crossover involving the basic steps for every evolutionary algorithms. The present work accentuates on an application of a genetic algorithm named strength pareto evolutionary algorithm (SPEA) for selection of features from crime datasets. The proposed work extracts crime reports from online newspapers providing crime information against women in Indian states and in its union territories. Each of the crime reports are encoded as bag-of-words and an exhaustive list of words have been prepared. Then the strength pareto evolutionary algorithm is used as a multi-objective optimization technique that provides a non dominated pareto front leading to selection of optimal features related to crime. The selected features also helps in further crime pattern analysis. This algorithm has two objective functions based on external clusters validation index and number of features in a sample. Significant research works related to optimization techniques exist in the past, but none have done global optimization on crime datasets. The proposed method gives better result than many other feature selection methods available.