An Efficient Machine Learning Approach for Crime Detection in India

Abstract

As the population is rising, levels of wealth are also reaching various social sections, and urbanization is also taking place in India, so crime is also changing unpredictably, which is making traditional ways of policing less efficient. Waiting for the occurrence of crimes to take place is not a good approach; rather, sensing a change in advance can be very helpful in catching crimes early, and in this case, the use of data technology such as pattern recognition is very effective. Comparing past crime statistics with population and income trends can help identify important relationships between events and their contexts. Rather than implementing a single approach, three distinct approaches are validated - Linear Regression, Random Forest Regressor, and Gradient Boosting Regressor - all of which are optimized by systematically adjusting their respective parameters. When comparing the results, the Gradient Boosting Regressor is able to provide more precise forecasts because it is able to interpret complicated relationships. This is especially true when comparing it to other models with similar characteristics. The new system, developed using Flask, functions within an efficient web tool. Crime predictions are provided to the police without any kind of delay. This allows them to modify their strategies since results are provided instantly. Their decisions become more refined since data is provided instantly.