Evaluation of Various Machine Learning Algorithms for Crime Prediction

Abstract

Crime poses a challenge to every nation's jurisdiction and administration. Thus, computerized crime forecasting and prediction may contribute toward making cities more secure. However, creating accurate and fast predictions about criminal activity is challenging. This is due to the incapability of humans that they can’t process large amounts of data and information. Thus, in the present scenario, machine learning algorithms are utilized in crime prediction models to analyse big data and find crime patterns based on various factors. In this paper, we have evaluated the various machine learning algorithms, namely, KNN, NN, RF, and NB for crime prediction. In the proposed model, the same dataset is trained and tested for different machine learning algorithms and find out which algorithm is effectively predicting the crime. In addition, pre-processing of the dataset is done to remove inconsistencies in the dataset and select the appropriate features using the correlation matrix. Further, the Chicago dataset is used for evaluation purposes and the code is designed and simulated with the help of Python and google colab software. Finally, the various performance metrics are determined for the crime prediction model and find out that NN outperforms over other machine learning algorithms.