Time Series Modelling of Daily Blood Donations in Malaysia

Abstract

Blood is a vital medical resource, yet maintaining a stable blood supply remains challenging due to fluctuating donation patterns, seasonal effects, and the short shelf life of blood products. These challenges were further highlighted during the COVID-19 pandemic, emphasizing the need for accurate forecasting to support effective blood supply chain management. This study applies traditional time series forecasting techniques to analyse daily blood donation patterns in Malaysia using data from January 2019 to February 2025. The study aims to apply the decomposition method to identify trends, seasonal patterns, and irregular fluctuations in daily blood donation data, and to compare the forecast performance of the Additive Decomposition method, Seasonal Autoregressive Integrated Moving Average (SARIMA), and Time Series Regression. Forecast accuracy is evaluated using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE) and Mean Absolute Percentage Error (MAPE). The results reveal clear weekly seasonality and moderate variability in donation patterns, with an overall increasing trend over time. Among the methods examined, Time Series Regression provides the most accurate forecasts with the lowest MAPE value of 36.69989092. The findings highlight the importance of reliable daily forecasting in supporting effective blood inventory planning and reducing shortages and wastage in Malaysia.