A Hybrid Explainable Deep Learning Framework for Multi Temporal Inventory Demand Forecasting and Decision Support

Modern supply chain management relies critically on accurate inventory forecasting to minimize
holding costs and prevent stockouts. However, traditional predictive models often struggle with
complex, multivariate demand patterns, and state-of-the-art deep learning approaches frequently
suffer from a "black-box" nature, limiting their adoption in operational decision-making. To
address this gap, this project introduces "Smart Inventory, " a hybrid Decision Intelligence framework that integrates deep learning-based multivariate time-series forecasting with Large Language Model (LLM)
driven explainability. The core predictive engine utilizes Poarch-based Long Short-Term
Memory (LSTM) neural networks to forecast future demand across multiple variables.
To bridge the gap between complex mathematical outputs and human operations, the
system employs an LLM integration to translate numerical forecasting thresholds into actionable,
human-readable purchasing insights. The framework is deployed as a full-stack architecture,
featuring a highly responsive Next.js frontend dashboard and a scalable Python Fast API backend.
Furthermore, the system incorporates a resilient, automated SMTP alert mechanism for proactive
low-stock management, engineered with robust fallback protocols for cloud deployment
constraints. Ultimately, this framework provides a comprehensive decision support system that not
only delivers high-fidelity demand predictions but also fosters user trust through Explainable AI.