Time Series Analysis and Crime Pattern Forecasting of City Crime Data

Abstract

Crime analysis using data mining techniques have been a possible solution to aid law enforcement officers to mitigate crime related problems. In this paper, a geospatial data analysis was conducted for detecting the hotspots of criminal activities in Manila City, Philippines. The crime records of 2012-2016 which were manually collected were geocoded and the map was generated using ArcGIS version 10. Association rules mining using Apriori algorithm was also performed on discovering frequent patterns to help the police officers to form a preventive action. This analyzed the different crimes and predicted the chance of each crime that can recur. In addition, analysis of various time series forecasting methods such as Linear Regression, Gaussian Processes, Multilayer Perceptron, and SMOreg to predict future trends of crime was performed. This work provides a solution to help the officers to build a crime controlling strategy to prevent crimes in the future.