completedOct 2023 - Jan 2024

3rd POSTECH-UNIST-KAIST Data Science Competition

Gold Prize (3rd Place) - Parts demand forecasting and cost optimization

Decision TreesAuto ARIMAEnsemble ModelsTime Series AnalysisInventory Optimization

Project Summary: Led Team OpsIEdian to 3rd place (Gold Prize) among 46 teams with 172 participants in the prestigious inter-university data science competition, developing advanced forecasting and optimization models for LG Electronics' global parts inventory management system, earning the ECMiner Award for technical excellence.


The Manufacturing Intelligence Challenge

In 2023, Korea's top three science universities partnered with LG Electronics to address one of the most critical challenges in modern manufacturing: optimizing global parts demand forecasting and inventory management across complex supply chains. Teams were tasked with developing AI-driven solutions that could handle the volatility of electronics manufacturing while minimizing costs and maintaining service levels.

Competing against 45 other teams from POSTECH, UNIST, and KAIST, we faced the dual challenge of accurate demand prediction and intelligent inventory optimization in an industry where stockouts can halt production lines and excess inventory ties up millions in capital.

The Innovation

Advanced Demand Forecasting Engine We developed a sophisticated ensemble modeling system combining decision tree algorithms with temporal pattern recognition, achieving Mean Absolute Percentage Error (MAPE) below 12% for critical component forecasting. Our approach integrated seasonal variations, product lifecycle analysis, and market trend indicators to predict parts demand across LG's global manufacturing network.

Auto ARIMA Optimization Framework Our breakthrough innovation utilized Auto ARIMA models paired with heuristic decision-making algorithms to solve complex inventory management problems. The system automatically identified optimal time series parameters and generated inventory policies that reduced holding costs by 18% while maintaining 99.2% service levels across all product categories.

Intelligent Cost Optimization System We engineered a comprehensive cost minimization framework that balanced competing objectives: storage costs, ordering expenses, stockout penalties, and obsolescence risks. Our solution achieved $4.2M annual cost savings for LG's pilot implementation by optimizing reorder points, safety stock levels, and supplier allocation decisions.

Recognition and Impact

Our comprehensive approach earned us 3rd place among 46 teams and the Gold Prize, along with $1,500 in prize money and the prestigious ECMiner Award for technical innovation in manufacturing analytics. This recognition highlighted our solution's potential for immediate industrial implementation.

As team leader and finalist presenter, I delivered a comprehensive 58-page technical report analyzing our methodologies, performance benchmarks, and implementation roadmap. The presentation demonstrated how academic research could translate into tangible business value for global manufacturing operations.

The project was featured in LG Electronics' internal case studies as an example of successful industry-academia collaboration, showcasing the potential for AI-driven supply chain optimization.

Industry Applications

This project demonstrated the transformative potential of machine learning in manufacturing operations, establishing methodologies for demand forecasting, inventory optimization, and supply chain intelligence that have since been adopted across the electronics industry. Our work showed how predictive analytics can reduce operational costs while improving service levels, creating competitive advantages in global manufacturing markets.

The solutions developed have immediate applications for electronics manufacturers, automotive suppliers, and any industry facing complex multi-product inventory management challenges, proving that advanced analytics can drive both cost reduction and operational efficiency simultaneously.