Scholarlink

Blood is an essential and rare commodity that can only be donated and used by humans. The need for blood products to be transfused to patients in hospitals is always high and urgent. Meanwhile, on the blood bank side, the supply from blood donations is uncertain, and the inventory depends greatly on the characteristics of different types of blood products. Therefore, this study aims to model blood product order demand forecasting from hospitals to support decision-making for many stages of the blood supply chain. The hybrid blood prediction model includes two prediction techniques, Gaussian process regression (GPR) and seasonal autoregressive integrated moving average (SARIMA), which are utilized to formulate forecasting models of the blood demand by processing highly complex data via GPR first and then using SARIMA by addressing both nonlinearities and seasonality in the data. The results with a mean absolute percentage error (MAPE) of 2.72% indicate that the hybrid model using GPR–SARIMA outperforms single GPR and other prediction methods and provides uncertainty estimates that bring benefits for decision-making about the range of blood demand.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Gaussian Process Regression and Seasonal Autoregressive Integrated Moving Average Approach for Dynamic Hybrid Blood Prediction Model

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Nguyen Le-Tuong,
  • Nguyen Van-Hop

摘要

Blood is an essential and rare commodity that can only be donated and used by humans. The need for blood products to be transfused to patients in hospitals is always high and urgent. Meanwhile, on the blood bank side, the supply from blood donations is uncertain, and the inventory depends greatly on the characteristics of different types of blood products. Therefore, this study aims to model blood product order demand forecasting from hospitals to support decision-making for many stages of the blood supply chain. The hybrid blood prediction model includes two prediction techniques, Gaussian process regression (GPR) and seasonal autoregressive integrated moving average (SARIMA), which are utilized to formulate forecasting models of the blood demand by processing highly complex data via GPR first and then using SARIMA by addressing both nonlinearities and seasonality in the data. The results with a mean absolute percentage error (MAPE) of 2.72% indicate that the hybrid model using GPR–SARIMA outperforms single GPR and other prediction methods and provides uncertainty estimates that bring benefits for decision-making about the range of blood demand.