Forecasting Food Donations in Saudi Arabia: A Machine Learning Case Study for Zero-Storage Food Banks

This study investigates how predictive analytics and data auditing can enhance the operational efficiency of a zero-storage food bank in Saudi Arabia. Drawing on historical data from Eta’am food bank across three cities, the research explores seasonal donation trends, donor return behavior, and data quality issues. Machine learning models including Random Forest, Linear Regression, ARIMA, and LSTM were applied to forecast donation volumes and classify donor return likelihood, with features incorporating Gregorian and Hijri calendar attributes. Results indicate strong seasonality linked to religious events, particularly Ramadan, and regional variation in donation behavior over weekends. Random Forest achieved high accuracy (98.85%) in predicting donor return, while Linear Regression outperformed more complex models in forecasting donation volume. A post-model data audit identified key data gaps, such as inconsistent calendar labeling and missing operational fields, leading to practical recommendations for data improvement. This study is one of the first to apply predictive modeling and auditing in a non-Western, zero-storage context and offers a replicable framework that supports both real-time logistics and strategic planning in alignment with Saudi Arabia’s Vision 2030.