Abstract: A process of drug shortage prediction includes defining a classifier and loading a listing of drug shortage events at different times in association with different drugs. Over different periods of time, transactions between nodes of the pharmaceutical supply chain are read on a data bus of a digital supply network computing platform, each indicating a receipt of a particular drug by a particular dispensary. Drug shortage events are then correlated with ones of the transactions at a specified duration of time preceding the drug shortage events. Consequently, the classifier can be trained with the correlated transactions annotated as giving rise to the drug shortage events. Finally, the classifier receives queries with newly observed transactions on the data bus and, in response, a likelihood is determined that a drug shortage event for a specific one of the drugs will occur at the specified duration of time following the newly observed transactions.