Enhancing Demand Forecasting for Long Lead Items using Bayesian Deep Learning: Driving Competitive edge theough Data Driven Inventory Decisions

This document is a doctoral research prospectus submitted as part of the Doctor of Business Administration (DBA) program at the Edgewood College. The study proposes a novel forecasting framework for long lead-time, low-volume finished goods, using Bayesian Deep Learning (BDL) techniques to model demand uncertainty. In addition, it introduces a proprietary accuracy metric—Floor-adjusted Weighted Mean Absolute Error (FWMAE)—to evaluate point forecasts in business-critical scenarios. The research aligns statistical forecasting with real-world inventory decision-making, supported by probabilistic evaluation metrics such as CRPS and PICP. This working paper lays the foundation for further empirical validation and industry application.