應用空間機器學習之緊急救護案件需求量預測分析

Abstract

將具緊急救護需求的病人，及時的送達適切的醫院，是緊急醫療救護服務運作的重要目標，若能增進運作效率，則可減低重創傷病患 (Major Trauma) 的死亡數，並提升傷病患的存活率。本研究應用支持向量機(Support Vector Machine)，類神經網路(Artificial Neural Network)，迴歸分析(Regression)及移動平均法(Moving Average)，建立緊急救護需求量之預測模式。預測結果可作為決策者派遣救護車或佈署醫療資源之參考依據，並期望增進到院前之緊急醫療服務效率。本研究以人口分布、醫療資源相對不均的新北市為研究案例，採用三年之緊急救護資料，進行每三小時與每日之需求量預測。為了更加了解緊急救護案件的時空間特性，本研究引入地理資訊系統 (Geographic Information System)，幫助建構資料管理模組，並透過彈性設定欲分析之時間與空間大小，對資料的時空間特性做分類與統計，並提升本研究應用於其他地區的可能性。最後，透過本研究所建構之資料管理與預測模組，初步成果顯示在本研究所建議之架構下，能提升預測績效，並有其潛力應用於實務中。

The objective for Emergency Medical Services (EMSs) is to deliver patients at the right place and time with the shortest response time. By increasing the operational efficiency, the survival rate of patients could potentially be increased. The Geographic Information System (GIS) is introduced to manage and visualize the spatial distribution of training data and forecasting results. A flexible model is implemented in GIS, through which training data are prepared with user desired sizes for spatial grid and temporal steps. The authors applied Moving Average, Artificial Neural Network, Regression, and Support Vector Regression for the forecasting of pre-hospital emergency medical demand. The results from these approaches, as a reference, could be used for the pre-allocation of ambulances. A case study is conducted for the EMS in New Taipei City, where pre-hospital EMS data has been collected for 3 years. With the easy use of the model and acceptable prediction performance, the proposed approach has been shown to have its potential to be applied to the current practice.