Machine Learning Algorithms for Edge-Assisted Crime Prediction and Evaluation

Abstract

The optimization and management of public safety has become more complex as a result of the new challenges posed by the world's expanding urban populations. Consequently, this study presents a method for forecasting a city's crime rate using demographic data and past events. Specifically, this research suggests a paradigm for evaluating and predicting criminal activity using edge machine learning techniques. Therefore, the effectiveness of the suggested framework is confirmed by a comprehensive examination of four separate types of crimes: drugs, repression of women and children, speedy trial, and murder. A graphic picture of crime in different locations of the nation has been presented by the whole research and implementation process. Choosing, evaluating, and deploying the Machine Learning (ML) model, followed by proposing the crime prediction, concludes the whole job. The use of categorization models allows for the prediction of criminal risk for a given time period and location. Decision Trees, Neural Networks, K-Nearest Neighbors, and Impact Learning are some of the ML approaches being used for event prediction, and their performance is being compared according to the data processing and modification employed. When it comes to crime prediction, the Decision Tree algorithm achieves an accuracy of up to 81%. The results show that using Machine Learning methods helps with crime event prediction, which has helped with public safety. Index Terms—DL, Decision Tree, KNN, MLP, Crime Prediction, Edge Computing, Impact Learning