DiffCrime: A Multimodal Conditional Diffusion Model for Crime Risk Map Inference
Published in Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2024
Recommended citation: Shuliang Wang, Xinyu Pan, Sijie Ruan, Haoyu Han, Ziyu Wang, Hanning Yuan, Jiabao Zhu, and Qi Li. 2024. DiffCrime: A Multimodal Conditional Diffusion Model for Crime Risk Map Inference. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD '24). Association for Computing Machinery, New York, NY, USA, 3212–3221. https://doi.org/10.1145/3637528.3671843 https://doi.org/10.1145/3637528.3671843
DiffCrime utilizes a multimodal diffusion model to integrate historical cases, satellite imagery, and map data, and accurately generates urban crime risk maps through HamNet. The Root Mean Squared Error (RMSE) of the two real-world datasets is reduced by 43% and 31%, respectively.
Recommended citation: Shuliang Wang, Xinyu Pan, Sijie Ruan, Haoyu Han, Ziyu Wang, Hanning Yuan, Jiabao Zhu, and Qi Li. 2024. DiffCrime: A Multimodal Conditional Diffusion Model for Crime Risk Map Inference. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD ‘24). Association for Computing Machinery, New York, NY, USA, 3212–3221. https://doi.org/10.1145/3637528.3671843