Data-Centric AI For Healthcare Supply Chain Forecasting

Abstract

Healthcare supply chains constantly experience the difficulty of demand forecasting especially in the events of disruption, which includes a pandemic. Conventional model dominated approaches to artificial intelligence (AI) and machine learning (ML) frequently face challenges of not keeping accurate predictions to such circumstances with inconsistent, incomplete, or poorly engineered inputs. The given research suggests an initiative of a data-centric AI approach to healthcare supply chain forecasting where the advanced data engineering practice is merged with ML and deep learning frameworks. The framework uses an extensible data pipeline that consumes heterogeneous information of procurement, consumption, logistics, and external information intake of epidemiological reports, mobility reports. Quality of data is checked using clear measures such as completeness, timeliness and representativeness and provide predictive models with credible inputs. The experiment conducted with the data on hospital networks proves that the addition of characteristics that have domain considerations to datasets can significantly enhance the accuracy of predictions. The data centric framework (over model centric baselines) improved the root mean squared error (RMSE) and stock-out rates by as much as 40% during disrupted demand conditions. These findings formulate the importance of data quality and engineering as key building blocks to resilient healthcare forecasting against earthquakes. The paper concludes that a more sustainable approach to finding a reliable adaptive supply chain process of supply chain intelligence implies looking beyond the concept of algorithmic optimization and to look instead at systematic data improvement.