Building a Production-Ready Time Series Forecasting System | FastAPI, XGBoost, LSTM
n this video, I demonstrate an end-to-end Time Series Forecasting System built for a production environment. The system automates the lifecycle of a machine learning model—from data preprocessing to model selection and API deployment.
Key Features:
Automated Model Benchmarking: Compares SARIMA, Facebook Prophet, XGBoost, and LSTM for 43 different states.
Smart Selection: Automatically selects the best-performing model based on validation MAPE.
Leakage Prevention: Implements strict walk-forward cross-validation logic.
Full-Stack Integration: A FastAPI backend serving predictions to a custom Next.js "Editorial-Terminal" dashboard.
Tech Stack: Python, FastAPI, TensorFlow (LSTM), XGBoost, Scikit-Learn, Next.js, Tailwind CSS.