Analysis of Weather, Temporal, Population, and Socio-Economic Factors in Determining Crime Rates in Five U.S. Cities and Projections for the Future

• Activity Number: 418
• Type: Topic Contributed
• Date/Time: Tuesday, August 11, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Scientific and Public Affairs Advisory Committee
• Author(s): Zhangxin Xue*
• Companies: Southern Methodist University
• Keywords: Regression ; Desicion Tree ; Netural Networks ; Ensemble ; SAS

Abstract

This presentation is based on the 2014 SAS Analytics and Data Mining Shootout. In this study, we examine the effects of weather, temporal, population density, and socioeconomic variables on crime rate by crime type by census tract within five U.S. cities: Atlanta, Chicago, Denver, Houston, and Sacramento. We were given 9 data sets including crime data, population data, unemployment data, storm data etc. We chose the hour as the major gradation of the data for our analysis. Data aggregation and imputation, preliminary bivariable analysis and data visualization were performed before modeling. Three modeling groups - regression, decision tree and neural network - were used, and best performing model from each group were combined to construct an ensemble model. Various results were provided, for each aspect of interests. Based on the results, recommendations on how the law enforcement resources can be more effectively utilized in reducing crime are made.