Machine Learning-Based Crime Pattern Analysis for Smart City Applications

Abstract - Due to an unprecedented rate of growth in the digital crime register within the urban zone, there is an essential need to address the different challenges faced by conventional theories of crime analysis. It has been witnessed that the manual process may not prove feasible enough for the analysis of large-scale multidimensional crime records in the context of smart cities. The aim of this research work is to design an efficient machine learning platform for crime analysis. The data that can be used is the data that is relevant to crimes, which gives information about the location and time at which the crimes are committed. The methods that are efficient during the data cleaning, normalization, and selection processes are used to improve the efficiency of the proposed system. Various machine learning algorithms, such as the Decision Tree classifier, Random Forest classifier, Naive Bayes classifier, and so on, are used comparatively. These systems are trained and tested on a well-structured experimental setup to analyse the predictive efficiency of the systems. The performance of the systems is analysed using different performance measures such as accuracy, precision, recall, and f1- score. Experiments were conducted to analyse the efficiency of different approaches using the ensemble method for predicting the crime pattern, where Random Forest is better than all other approaches. The findings show that machine learning methods have the potential to effectively aid intelligent crime analysis systems as well as smart cities in their provision of proactive policing, crime hot spot detection, and optimal allocation.