3505. Using crime report app data to predict burglary via machine learning
Invited abstract in session MB-31: Crime Analytics, stream Analytics.
Monday, 10:30-12:00Room: 046 (building: 208)
Authors (first author is the speaker)
| 1. | Sebastian Maldonado
|
| Department of Management Control and Information Systems, University of Chile | |
| 2. | Joaquín Roa
|
| Centre for Public Systems, University of Chile | |
| 3. | Carla Vairetti
|
| Universidad de los Andes | |
| 4. | Richard Weber
|
| Department of Industrial Engineering, FCFM, University of Chile |
Abstract
Predictive policing represents a productive avenue of research utilizing analytics to anticipate potential criminal activities. This study aims to develop a learning machine capable of forecasting the occurrence of specific crime types in a given area based on historical data. The primary objective of this model is to furnish municipalities and police departments with a tool to deploy patrols and allocate resources efficiently. Data was collected from a Chilean crime report app, encompassing various criminal events and suspicious activities reported by the public. Under the assumption that past reports serve as reliable predictors of future criminal activities, we formulate a classification problem to forecast burglary and motor vehicle theft in a designated area. Promising outcomes were achieved using conventional statistical and machine learning methods such as logistic regression, decision trees, and gradient boosting, attaining a balanced accuracy exceeding 80%. We also discuss the used of graph-based deep learning and other network architectures to take advantage of the spatial structure of the problem.
Keywords
- Artificial Intelligence
- Analytics and Data Science
- Critical Decision Making
Status: accepted
Back to the list of papers