Cluster Detection for Traffic Accidents on Spatiotemporal Networks

This presentation explores a method for detecting high-risk locations and times for traffic accidents using historical data. This work was presented at KES2024.

Transcript

1. 28th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Seville, Spain, 11-13 September 2024 Cluster Detection for Traffic Accidents on Spatiotemporal Networks Kesiuke ANDOa, Yusuke KUNIYOSHIb, Natsuki ONOGIa, Takeshi UCHITANEa Naoto MUKAIc, Kazunori IWATAd, Nobuhiro ITOa, Yong JIANGe a Aichi Institute of Technology, 1247 Yachigusa, Yakusa-cho, Toyota-shi, Aichi 470-0392, Japan b The University of Electro-Communications, 1-5-1 Chofugaoka, Chofu-shi, Tokyo 182-8585, Japan c Sugiyama Jogakuen University, 17-3 Hoshigaoka-motomachi, Nagoya-shi, Aichi 464-8662, Japan d Aichi University, 4-60-6 Hiraike-cho, Nakamura-ku, Nagoya-shi, Aichi 453-8777, Japan e Aichi University, 1-1 Machihata-cho, Toyohashi-shi, Aichi 441-8522, Japan 28th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Seville, Spain, 11-13 September 2024
2. 28th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Seville, Spain, 11-13 September 2024 TRAFFIC SAFETY IN JAPAN Governments implement various traffic safety measures: Toyota Transportation Research Institute, Safety map (https://www.safety-map.info/) Aichi Prefectural Government, Autumn National Traffic Safety Campaign (https://www.pref.aichi.jp/soshiki/shinshiroshitara /0000076401.html) Traffic safety campaign Safety map (“Hiyari” map)
3. 28th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Seville, Spain, 11-13 September 2024 WHERE TO FOCUS FOR TRAFFIC SAFETY ?
4. 28th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Seville, Spain, 11-13 September 2024 ACCIDENT CLUSTER DETECTION IN OUR STUDY
5. 28th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Seville, Spain, 11-13 September 2024 CHALLENGES IN OUR STUDY
6. 28th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Seville, Spain, 11-13 September 2024 OUR OBJECTIVES IN THIS STUDY Enhance reliability • Develop a method to select data that satisfies the model assumptions; • Evaluate the accuracy of the clusters using metrics; Enhance generalizability • Apply our method to data from multiple regions and assess its effectiveness.
7. 28th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Seville, Spain, 11-13 September 2024 ACCIDENT CLUSTER DETECTION
8. 28th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Seville, Spain, 11-13 September 2024 OUR APPROACH
9. 28th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Seville, Spain, 11-13 September 2024 DATA SELECTION BASED ON HYPOTHEIS TESTING
10. 28th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Seville, Spain, 11-13 September 2024 METRIC: PEDICTIVE ACCURACY INDEX (PAI)
11. 28th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Seville, Spain, 11-13 September 2024 TARGET REGIONS • Location: 8 cities in Aichi Prefecture, Japan; Period: 2015 – 2020; • Area: 6.5km (east-west) x 4.5 km (north-south); • Each data record has latitude, longitude, and time of occurrence. Maptiles by MIERUNE, under CC BY. Data by OpenStreetMap contributors, under ODbL.
12. 28th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Seville, Spain, 11-13 September 2024 DATA SATISFIED PRECONDITIONS
13. 28th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Seville, Spain, 11-13 September 2024 ACCURACY OF TRAFFIC ACCIDENT CLUSTERS
14. 28th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Seville, Spain, 11-13 September 2024 INTERPRETING DETECTED CLUSTERS
15. 28th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Seville, Spain, 11-13 September 2024 CONCLUSION Improving reliability and generalizability: • Developed a method to select data that satisfies the model's assumptions; • Evaluated cluster accuracy using PAI; • Validated the effectiveness of our method across diverse regions. Results: Our model consistently outperforms previous methods in all regions.
16. 28th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Seville, Spain, 11-13 September 2024 ACKNOLWEDGEMENTS • We are grateful to K. Shimizu for valuable discussions. • Special thanks to the Aichi Prefectural Police for providing traffic accident data. • Supported by the “Agent-Mediated Driving Assistance Research Project“ of the MHI and Human Characteristics Research Division, Nagoya University. • Supported by the “Fatal Accident Data Analysis in Aichi Prefecture using AI and GIS“ joint research project of the Institute of Managerial Research, Aichi University. • Funded by JSPS KAKENHI Grant Number 23K11225
17. 28th International Conference on Knowledge-Based and Intelligent Information & Engineering Systems Seville, Spain, 11-13 September 2024 REFERENCES [1] Ando, K., Kuniyoshi, Y., Shimizu, K., Uchitane, T., Mukai, N., Iwata, K., Ito, N., Jiang, Y., 2023. Spatio-temporal Network Analysis for Detecting Traffic Accident Clusters, in: 2023 14th IIAI International Congress on Advanced Applied Informatics (IIAI-AAI), pp. 257–262. [2] Xie, Z., Yan, J., 2013. Detecting traffic accident clusters with network kernel density estimation and local spatial statistics: An integrated approach. Journal of Transport Geography 31, 64–71. [3] Romano, B., Jiang, Z., 2017. Visualizing Traffic Accident Hotspots Based on Spatial- Temporal Network Kernel Density Estimation, in: Proceedings of the 25th ACM SIGSPATIAL International Conference on in Geographic Information Systems, pp. 1–4. [4] Anselin, L., 2005. Exploring Spatial Data with GeoDa: A Workbook. Urbana 51, 1–244 [5] Chainey, S., Tompson, L., Uhlig, S., 2008. The Utility of Hotspot Mapping for Predicting Spatial Patterns of Crime. Security Journal 21, 4–28.