Analysis & forecasting of juvenile crime using variance threshold and time series algorithm

Juvenile crime, a prevalent issue in the contemporary world, encompasses unlawful activities committed by individuals below the age of 18. Offenses such as theft, assault, burglary, rape, vandalism, and various other criminal acts fall within this category. In India, the escalation of juvenile criminal behavior is noteworthy, prompting the need for advanced techniques and algorithms. This research explores into the models, fostering data-driven decision-making. Policymakers can leverage insights from the study to formulate informed decisions regarding law enforcement strategies and the development of rehabilitation initiatives for juveniles. This research provides recommendations on selecting the appropriate forecasting model based on the age group of individuals and the type of crime to be predicted. In this paper data has been referred from NCRB Publications on Crime in India “Juveniles Apprehended under IPC Crimes and SLL by Age Groups and Sex”. Out of 55 IPC crimes, 34 were suitable for forecasting, with burglary, rape, hurt, murder, and theft identified as the top five most frequent crimes, and data from these crimes over the past 12 years has been used for modelling and forecasting. The result after Leave out one Validation shows that Improved manual ARIMA gives better results than Auto ARIMA in all the three age categories with top five crimes based on the Akaike's Information Criterion and Evaluation metrics.