Crime Data Forecasting using Exponential Smoothing

Abstract

Crime data analysis to predict the crime that is likely to happen in the future, can be established using mathematical models. Such forecasting is done through data mining techniques to extract relevant information and reveal patterns from existing sets of data. This study investigated forecasting models using Exponential Smoothing (ES) for crime incidents geared towards the development of a resource allocation recommendation system. Comparison of the ES models using different alpha values from 0.2 to 0.7 were subjected to forecast accuracy test using Mean Absolute Deviation (MAD). The Quantitative Forecasting method was used on the dataset which consisted of the crime reports recorded from 2016 to 2018 from the six (6) Angeles City Police Stations (ACPS) under Angeles City Police Office (ACPO) in Region 3. The application of data cleaning process allowed the researchers to pinpoint outliers from the crime data and the crime incidents, summarized according to types of crimes committed within the jurisdiction of the police stations in the locale. The Exponential Smoothing with various models using different alpha values was used. Results showed that such method used is significant in the monthly analysis of crime data. Each of the crime incident types indicate the need for a different alpha value to be used. The police station’s resource allocation recommendation system with crime incidents forecasting was made using prototyping methodology. The recommendation feature can help the ACPO in planning its allocation of resources for more efficient logistic management.