Inventory-Aware Return and Stockout Prediction for U.S. Retail E-Commerce - Joseph Tran

Inventory-Aware Return and Stockout Prediction for U.S. Retail E-Commerce: A Machine Learning Approach Course: DS687 Description: This capstone outlines a methodology for a machine learning framework that regards two costly setbacks in online retail: product returns and inventory shortages as parts of the inventory risk. It has two models based on a common, interpretable pipeline: a return risk classification model (logistic regression and random forests with ModCloth Fit data) and a stock-out risk classification model (XGBoost with time-based validation and Walmart M5 data). Both models’ outputs are rolled out into a rule-based decision layer developed in the form of an interactive Streamlit web application called “Inventory Risk AI”.