FortiRx 2.0: Smart Privacy-Preserved Demand Forecasting of Prescription Drugs in Healthcare-CPS

Pharmaceutical Supply Chains (PSCs) are a combination of processes and networks involved in the production, distribution, and delivery of pharmaceutical products from the stage of raw materials till they reach the end-user which is typically patients. PSC is a complex and critical part of the Healthcare Cyber-Physical System (H-CPS) that ensures the availability, quality, and timely delivery of pharmaceutical products to meet the needs of the patients. One of the main characteristics of efficient PSC is accurate production planning and resource allocation which not only reduces the costs but also significantly enhances customer satisfaction. Accurate production planning for manufacturers needs accurate demand prediction models in place, otherwise can lead to overstocking or under-stocking, wastage, and other costs. Demand prediction of pharmaceutical products is very complicated due to unexpected fluctuations due to seasonality, pandemics, and several other reasons. This causes supply chain disruptions which may in turn cause both life and financial losses. Many statistical methods and machine learning models including time series analysis for forecasting the demand based on historical, but the significant setback is the availability of real-time data and privacy concerns on prescription data being shared. The current proposed FortiRx 2.0 architecture leverages a blockchain-based prescription system and federated learning approach to solve these issues and provide accurate demand forecasting. Federated XGBoost is used for demand prediction and different evaluation metrics are computed for performance evaluation of implemented model. These metrics are compared with baseline models Naive and Seasonal Naive as a reference and the results from the comparison are discussed.