Enhancing Crowding Event Detection on Campus with Multidimensional Logs: A Meta-Heuristic Search Approach

Maoyi Wang, Jiajie Shen, Jack Mao, Jihan Dai, Bochun Wu, Yun Xiong, Xin Wang

Abstract: This research addresses the pressing need for rapid detection of emergencies in campus crowding events by introducing wireless log-based crowding detection (WLCD). The WLCD employs a novel meta-heuristic search algorithm, coupled with the utilization of wireless fidelity (Wi-Fi) network logs, hierarchical attribute relationships, and information entropy theory. As the first attempt to transform a crowding event detection problem into a combinatorial optimization problem, we validated our approach using a real-world data set of 500 million records from a campus Wi-Fi network with over 27,000 access points. Our study demonstrates the superior performance of the WLCD, with increased accuracy by up to 61% and reduced execution time by up to 82% compared to state-of-the-art methods.

External IDs:dblp:conf/cocoon/WangSMDWXW24