AI-Driven Dynamic Reorder Point Prediction for Inventory Management in Supply Chains: A Machine Learning Approach

Abstract

Keeping the right amount of stock in a modern supply chain is a tricky balancing act. It comes down to making sure products are there when people want them without spending a fortune on warehouse space, all while demand shifts and changes in ways that are hard to pin down. Most places still use old-school reorder point methods. These work fine enough, but they usually bank on everything staying the same, which just isn't how retail works anymore. This research looks into how AI might handle those shifting numbers better by creating dynamic reorder points based on actual retail transaction data. It involved building a specific pipeline to turn messy sales records into clear demand patterns for individual items, then testing out various machine learning models to see if they could actually predict what happens next. Those predictions were then plugged into a system designed to adjust reorder levels on the fly and tested in a simulated environment. The numbers show that machine learning does a much better job of guessing demand than the basic statistical shortcuts people used to rely on. When these models were used to set reorder points, the system got much better at preventing empty shelves and keeping customers happy. But there is a catch. Better service usually meant holding onto more stock, which points to a clear tension between being reliable and being cheap. Looking closer at the data, it became obvious that how long it takes for a shipment to arrive really changes how well these policies work. Plus, even the smartest models struggled when demand went totally off the rails or when weird, one-off transactions popped up. In the end, it seems that using Machine Learning to move reorder points around can actually help a business run better, provided the system is set up thoughtfully. This work offers a clear, repeatable way to link demand guesses with inventory simulations, giving a solid starting point for anyone trying to build better decision tools for supply chains. It also serves as a reminder that a more accurate forecast only matters if it actually leads to a smarter decision on the warehouse floor.