Open Source Time Series Forecasting System
A comprehensive MLOps solution that automates the entire demand forecasting workflow from data ingestion to model deployment. The system supports multiple forecasting algorithms and automatically selects the best performing model for each product category.
Problem Statement
Retail businesses struggle with accurate demand forecasting, leading to overstock or stockouts. Manual forecasting processes are time-consuming and don't scale well across thousands of SKUs.
My Approach
Designed a modular pipeline architecture using Apache Airflow that handles data preprocessing, feature engineering, model training, and prediction generation. Implemented automated model selection based on historical accuracy metrics.
Key Outcomes
- Automated daily forecasting for thousands of SKUs
- Model accuracy improvement through ensemble methods
- Reduced manual intervention in forecasting workflow
- Scalable architecture supporting multiple clients
Screenshots
Tech Stack
Tags
Project Info
- Status
- In Development
- Category
- AI-ML
- Created
- 1 year ago
- Ended
- Present
Related Projects
AI Demand Forecasting Development at Sony
Developed the Focus model within a toolkit used for AI-driven demand forecasting on Azure Databricks platform at Sony.