GAM and Time Series Analysis based Model for Predicting the Blood Donation

• Main Authors: You-Chen - Chen, 陳宥辰
• Other Authors: Shi-Woei Lin
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/59098726270535228809

碩士 === 國立臺灣科技大學 === 工業管理系 === 105 === Blood is a vitally important fluid of the human body, and the management of its supply and demand is an important research issue in contemporary health care systems. Because blood substitute has not yet been well developed, most of the healthcare systems rely on voluntary donations to ensure sufficient blood supplies. However, blood demand and supply can be highly stochastic and irregular because the donation pattern and transfusion need may be affected by various factors including time, weather, and other events (such as disasters). Blood products are also perishable, and an inappropriate inventory management can result in shortage or wastage of blood products. Thus, a model which is capable of accurately predicting blood supply and demand has long been valued in the blood inventory management research. In this study, a forecasting model which can precisely predict the amount of blood donation was developed. In particular, we integrated the data of donation histories of whole-blood donors collected from Taipei Blood Center and the weather information form the Central Weather Bureau, and combined the generalized additive model (GAM) with the time series model to construct a two-step prediction model. A bootstrap method was also implemented on the residuals produced by the two-step model to construct the prediction intervals which can later be used to detect the anomaly donation patterns. The results of this research could have considerable implications in blood inventory management to help achieving better allocation of resources for Blood Centers.