Spatiotemporal Bandits Crime Prediction from Web News Archives Analysis

Abstract

It is said that prevention is better than cure. Hence the idea of preventing crime from occurring is the best for public safety. This can only be achieved if the law enforcement agencies have a prior knowledge of where and when a crime will occur. A crime is an act that is criminal under the law. It is detrimental to society to comprehend crime in order to prevent criminal action. In order to prevent and solve crime, data-driven research is beneficial. Bandit crime has been on the rise in Nigeria, thereby causing public disorder. In this study, from the perspective of artificial intelligence, a novel hybrid deep learning model for crime prediction is proposed. Bandits’ crime datasets are obtained online through news archives which are less expensive. Spatial crime analysis was carried out on the novel bandit crime dataset obtained and prediction were made using the newly proposed DECrimeXGBoost model. A comparative analysis was performed with respect to precision, recall, f-measure, and accuracy with other crime predictions algorithms and the proposed model outperformed the other algorithms with accuracy of 99.9999%.