Fast Algorithms for Three-State Spatial Queuing Problems with Applications to Large Scale Emergency Service Systems
时间：2019年12月3日（周二）10:00-11:00
地点：浙江大学紫金港校区行政楼1002会议室
主讲人：花成 博士生，耶鲁大学商学院
主持人：陈熹 教授，浙大yl23411永利
摘要:
We focus on modeling and evaluating a novel emergency service system in which cross-trained fire-medics respond to both fire calls and medical emergencies. Fire demand in the U.S. has decreased dramatically in the last three decades, while emergency medical calls have surged. With this changing landscape, cities are under pressure to reduce their budgets by closing fire stations. We show that a better alternative would be to implement a fire-medic system in terms of cost savings and response time performance. We develop an exact spatial queuing model and two approximation methods, the second of which has linear complexity and can be used to solve three-state problems of any size. This paper is the first to develop a fast algorithm for general three-state spatial queuing problems. In several constructed examples, performance errors are less than 1% compared to exact values. We apply our method to the fire-medic system in St. Paul, MN. and find the errors between predicted and actual average response times are less than 2%. We show, in St. Paul, the traditional system would require 33% more personnel than the fire-medic system to achieve the same mean response times. We believe the fire-medic approach and our modeling have widespread applicability to other cities.
主讲人简介：
花成，耶鲁大学商学院运筹学博士生。他获得了耶鲁大学统计与数据科学系的硕士，上海交通大学计算机工程学士，美国密西根大学工业工程学士，以及数学二专学位。他的研究致力于大数据驱动下的优化算法、机器学习、深度学习与随机过程、队列理论的结合，在大数据城市医疗系统、量化金融、与体育分析中的应用。
欢迎广大师生前来参加！
