Prediksi Jumlah Permintaan Darah Jenis Packed Red Cells Menggunakan Support Vector Regression

Abstract

The Blood Transfusion Unit (UTD) of the Indonesian Red Cross (PMI) in Probolinggo Regency faces challenges in managing blood supplies, particularly for Packed Red Cells (PRC), due to unpredictable fluctuations in demand. The imbalance between supply and demand often leads to shortages or surpluses, affecting the quality of healthcare services. This study aims to predict the demand for PRC to support more efficient inventory management. The method employed is Support Vector Regression (SVR), an approach within the Support Vector Machine (SVM) algorithm that is effective for regression and prediction tasks. Historical blood demand data was used to train the model. The results indicate that SVR provides sufficiently accurate predictions for all blood types, with the best performance achieved for blood type B, yielding a Root Mean Square Error (RMSE) of 0.0589 and a Mean Absolute Percentage Error (MAPE) of 7.11%. In conclusion, the SVR method can be effectively applied to forecast PRC demand and has the potential to support decision-making in blood stock management at the UTD PMI of Probolinggo Regency.