Intelligent Automation of SFO Crime Prediction using Multiple Artificial Intelligence Methods

Crime needs a broad understanding of its patterns and reasons behind specific types that happen in certain areas only. The crime pattern theory provides the explanation of distribution of criminal events and its variations. Crime generators and crime attractor concentrate in the places that include frequent routine movements of the population to create hot and cold spots of crime. The offender is more often comfortable with frequently committed crimes and prefers to commit these in the places that they are most familiar with. These offenders can be distinguished based on the type of crimes that fall into specific incident categories. Identifying these categories help to categorize not only frequently occurring crime locations but also facilitate the police department to consider the type of support or action that needs to be planned. Special analysis of crimes and its categories are essential to understand the frequency, time and its patterns. This paper showcases the use of different Artificial Intelligence techniques and compares their behavior and outcomes which includes different machine learning and deep learning techniques like random forest, K-nearest neighbor, Artificial Neural Network, TabNet and Time Series. These are different flavors of Artificial intelligence and also interactive dashboards and web applications are supported to visualize the hidden patterns and in depth details the different features. The flexibility of the paper is extended to support any type of crime dataset with a minor initial data streamline process and complete end to end flow is built using only a python program to reduce the infrastructure cost. The benefit from this paper is for the police stations to plan for more staffing to avoid the occurrence of crimes and push towards prevention steps. It helps the policy maker to visualize and provide more knowledge on the crime occurrence and prediction.