Predictive Modelling of Crime Data using Machine Learning Models: A Case Study of Oyo State, Nigeria

Nigeria's criminal scene is characterized by violence, terrorism, alienation of citizens, and political instability. The Global Peace Index has listed Nigeria as one of the nations with the least amount of peace in the globe as of late. It is the nineteenth least tranquil condition. Nigeria is also ranked as the ninth most terrorist-affected country in the world by the Global Terrorism Index. The potential for genocide, or mass murder, is another grave hazard facing Nigeria. Nigeria was the fifth-highest risk country in Africa and the twelfth-highest risk country globally as of the end of 2023. As a result, it is necessary to analyze Nigeria's crime data and create suitable modeling for projections in the future. This work examines the various crime data between 2013 to 2023 as reported by the Nigeria Police Force in Oyo State. Murder, rape, indecent assault, armed robbery, theft, burglary, assault, and kidnapping are crimes considered. To examine the connections between crimes and ascertain their distributions, correlation analysis was employed. In the thirty-three (33) local government areas of Oyo State, big crimes are predicted using machine learning techniques including Autoregressive Integrated Moving Average (ARIMA), Autoregressive Fractional Integrated Moving Average (ARFIMA), and Long Sensory Term Memory (LSTM) models. The result shows that, LSTM produced better performance in modelling crime data compared to ARIMA and ARFIMA models with lowest AIC values. Through this approach, the government and security services would be able to plan appropriately for the likelihood of future crimes in Oyo State.