Leveraging Parallel Spatio-Temporal Computing for Crime Analysis in Large Datasets: Analyzing Trends in Near-repeat Phenomenon of Crime in Cities
TimeTuesday, July 246:30pm - 8:30pm
DescriptionCrime often clusters in space and time. The pattern of clustering in time and space helps inform crime prevention measures and could improve prediction of future activities. Analytical methods for crime analyses help to identify these underlying spatio-temporal patterns, but must differentiate statistically significant clusters from random patterns. Previous research have shown that Monte Carlo simulations could be used to assess the statistical significance of these patterns. However, applying Monte Carlo simulations to numerous crime events raises computational challenges. In-order to tackle this challenge we develop a software based on shared memory parallel programming paradigm that could efficiently process large sets of spatio-temporal crime data to extract near repeat patterns and asses the statistical significance. We use parallel spatial domain decomposition strategy to speed-up the domain decomposition process and further utilize multiple processors to generate Knox table based on inter-distance and inter-time calculations. Different spatio-temporal configuration for sub-domains created after domain decomposition will be tested to estimate the configuration that could provide the best performance. For this study we utilize freely available crime datasets from four major cities including Washington DC (215K records), New York City (5.58M records), San Francisco (2.14M records) and Chicago (6.4M records). The new software that we propose for near-repeat modelling would be helpful for researchers as well as authorities to analyze large datasets of crime events.