Classification of Crimes Using Machine Learning Techniques for National Crime Data

Abstract

The paper aims to assist in a more accurate classification of crimes provided by North Macedonia’s Ministry of Internal Affairs using machine learning techniques. By means of natural language processing, data is transformed into a format suitable for term frequency-inverse document frequency (TF-IDF) vectorization. Bayesian and Support Vector Machine models are utilized and evaluated. The results show non-negligible improvement toward a successful classification of crimes, confirming the beneficial nature of machine learning techniques towards such tasks.