Abstract | ||
---|---|---|
Crime prediction has attracted increasing attention due to its significance in public safety and growing availability of heterogeneous relevant data. Existing works on crime prediction usually failed to capture its dynamics and inherent non-linear relationships. To address these issues, in this paper, we propose an attentional recurrent neural network for future crime count prediction leveraging heterogeneous urban open data. In particular, we first extract and embed relevant features using multi-source data, e.g. crime records, POIs, demographic data and meteorological data. We then feed all features into a two-layer recurrent neural network to capture temporal relevance. Further, we incorporate an attention mechanism to capture the time-varying dependency. The final prediction results can be generated through a fully connected neural network. Extensive experiments with real-world datasets verify the effectiveness of our proposed framework which outperforms many baseline methods. |
Year | DOI | Venue |
---|---|---|
2022 | 10.1007/s42486-022-00100-x | CCF Transactions on Pervasive Computing and Interaction |
Keywords | DocType | Volume |
Crime occurrences prediction, Heterogeneous data, Urban computing, Recurrent network | Journal | 4 |
Issue | ISSN | Citations |
3 | 2524-521X | 0 |
PageRank | References | Authors |
0.34 | 6 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Fangxun Zhou | 1 | 0 | 0.34 |
Binbin Zhou | 2 | 0 | 2.03 |
Sha Zhao | 3 | 48 | 9.96 |
Gang Pan | 4 | 1501 | 123.57 |