Examining Deep Learning Architectures for Crime Classification and Prediction
Problem
Predicting crime types and high-risk crime areas from open police-report time-series data per location.
Why it matters
Reliable crime forecasting could help police allocate resources and prevent crime.
What the AI did
Compared 3 deep learning configurations against 10 state-of-the-art methods for crime classification/prediction, and derived configuration recommendations for improved performance.
Results
Deep learning methods consistently outperformed existing best-performing methods across 5 public datasets; no numeric metrics stated.
Impact
Evidence that deep learning improves crime area prediction over existing methods on public data.
Record
| Venue | Forecasting |
|---|---|
| Published | 2021 |
| Work type | article |
| Authors | Panagiotis Stalidis, Theodoros Semertzidis, Petros Daras |
| Citations | 20 (OpenAlex, at import time) |
| References | 51 |
| Model or method | deep learning architectures (3 configurations; specifics not stated in source) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.3390/forecast3040046 |
| Open access | gold |
| Keywords | deep learning, crime prediction, time-series, classification, benchmarking |
Importance rating
real but narrower advance
Crime forecasting could aid police resource allocation and prevention; societal value is plausible but abstract states no scale or deployment evidence.