Improved neural network for predicting blood donations based on two emergent factors

Area of Science:

• Healthcare Management
• Supply Chain Logistics
• Predictive Analytics

Background:

• Effective blood supply chain management relies on accurate blood donation forecasting.
• Limited research exists on modeling blood donation influenced by emergency events.
• Understanding these influences is key to optimizing blood availability.

Purpose of the Study:

• To investigate the impact of diverse emergency events on blood donation patterns.
• To develop a robust blood volume prediction model incorporating emergency factors.

Main Methods:

• Utilized time series data of blood donations from 2015-2021.
• Employed statistical methods to analyze weather and epidemic influences.
• Developed a hybrid Seasonal Autoregressive Integrated Moving Average with eXogenous factors (SARIMAX) and neural network model.

Main Results:

• Identified significant variations in blood donations during emergencies.
• Confirmed a strong correlation between epidemic status and donation volumes.
• The hybrid model improved prediction accuracy by 60.7% (decision coefficient) and reduced Root Mean Square Error (RMSE) by 1.668.

Conclusions:

• The hybrid SARIMAX-neural network model enables accurate, long-term blood donation forecasting, even under emergency conditions.
• Provides reliable 2-week advance forecasts for blood stations.
• Facilitates improved hospital blood supply, treatment rates, and reduced transportation costs.