行動式中風診療單元部署之最佳化研究

Abstract

資源在空間上的分配已被探討許久，如緊急醫療服務、學校、警局等設施在地理空間的設置。現今大部分的討論都建立在「最大覆蓋位置問題」(Maximal Coverage Location Problem)，該模型旨在找到給定數量的設施在空間中的最佳放置位置的問題，目標是在一定限制下最大化覆蓋人口的總需求。此研究我們以找出移動式中風診療單元的部屬地點為目的，使用美國加利福尼亞州馬林縣的緊急救護資料，從機器學習的角度，用最大期望演算法(Expectation-maximization algorithm)找出最佳部屬地點。再輔以救護車派遣模擬估計派遣花費的時間。考慮到MCLP是一個NP-hard問題，EM衍生方法不需要複雜的數據預處理，能更快地提供解決方案，這使其成為解決覆蓋位置問題的一個更有效的替代方案。

The allocation of resources in spatial distribution has been a topic of discussion for a long time, including the placement of emergency medical services, schools, and police stations in geographic spaces. Most current discussions are based on the Maximal Coverage Location Problem (MCLP), a model focused on finding the optimal placement of a given number of facilities in a space to maximize the total demand coverage under certain constraints. In this study, we aim to determine the deployment locations for Mobile Stroke Units (MSUs) using emergency medical data from Marin County, California. From a machine learning perspective, we employ the Expectation-Maximization (EM) algorithm to identify the optimal deployment sites. Additionally, we use an ambulance dispatch simulation to estimate the dispatch time. Considering that MCLP is an NP-hard problem, the EM derivative method, which does not require complex data preprocessing, provides a quicker solution. This makes it a more effective alternative for solving the coverage location problem.