应用混合式人工智能模型于急诊人数预测.docx

PAGE II PAGE II PAGE II 中文摘要 在现今社会里，医疗服务业为目前社会中无法被取代的重要产业，而其中的紧急诊疗更为重要且急迫需要，紧急诊疗为俗称的急诊。但由于现今保健政策与费用逐日增加，加上医疗作业人员的薪资与劳力付出不成比例，形成了急诊人力经常性的短缺，有鉴于此对于急诊病患（Emergency Department visits, ED）的到诊时间的预测便显得十分重要。 对于现今服务业所使用的时间序列数据，经常性的无法采用单一预测模式进行运算，必须思考其特性以便进行分析，因此本研究将考虑气候状况与环境气温变化等外部因子，透过组合预测法（Combination Forecasting）结合倒传递神经网络（ Back Propagation Neural Network, BPN ）及自我回归移动平均 （Autoregressive Integrated Moving Average Model, ARIMA）之运算优势，提高预测之精准度。并将预测结果与一般 BPN 及 ARIMA 预测模型进行比较。 本研究将医院给予之急诊病患到诊时间切割为两小时，目的在于实用价值较高。过去研究显示外部因子具有一定之影响力，本研究透过组合预测结合 BPN 及 ARIMA 之优势，实时反应出急诊到诊人数突发性之变动。研究结果显示组合预测效果确实较为良好且稳定。 关键词：急诊病患、自我回归移动平均、倒传递神经网络、组合预测法 英文摘要 The medical service industry is one of the irreplaceable industries in the modern society, and within the scope of medical services, none is more pertinent and important than the emergency medical service (Emergency Service). However, the recent changes in health insurance policies and the rising costs of health insurance, coupled with disproportionate wage and labor contribution of the medical staff, have resulted in significant and frequent shortages of emergency medical manpower. As such, the prediction of emergency patients is increasingly important. Due to the inability to calculate time series data in the service industry with a single forecast model, the characteristics of the industry must be considered before any analysis is to be performed. Our study took into consideration external factors such as climate conditions and changes in environment temperatures, integrated with Back Propagation Neural Network (BPN) and Autoregressive Integrated Moving Average Model (ARIMA) through combined forecasting to improve the calculation deficiency of support vector regression, increasing the accuracy of our prediction. We then performed comparison of the predicted results with BPN and ARIMA. In the current study, we divided one day into twelve interval equally to correspond to the shift change of the emergency staff. Ou