A machine learning-based framework for predicting supply delay risk using big data technology

The growing complexity of products and business processes is pushing companies toward an integrated and data-driven supply chain management. Companies are increasingly adopting big data analytics (BDA) and machine learning (ML) approaches as a way to manage uncertainty factors and to soften their effects on supply chain performance. In this paper, we present a data-driven framework that combines BDA techniques and ML models for estimating the risk of supply delay at a multinational automotive electronics manufacturer. The framework is based on a big data architecture so as to facilitate the integration of the proposed models into real practical contexts driven by large volumes of data. The framework developed was empirically tested with real data. We evaluated the results obtained from the predictive models not only in terms of error metrics, but also in terms of the financial impact of misclassification on inventory management performance. Our results could be used to promote the adoption of modeling strategies that relax common assumptions in the supply chain literature, such as considering that supply lead time is constant or that its distribution is known.