SRM Datathon 2025
XYZ Ltd., a leading heavy machinery manufacturer, faces key operational hurdles:
- Demand Volatility: Fluctuating sales due to seasonal cycles, government budgets, and market dynamics.
- Inventory Imbalance: Overstocking raises holding costs; stockouts cause revenue loss and customer dissatisfaction.
- Storage Constraints: Fixed warehouse capacity of 5000 cubic meters demands precise stock planning.
We built an intelligent forecasting and inventory management system powered by advanced ML and optimization techniques.
Models Used:
XGBoost: Gradient boosting for tabular sales predictionLSTM: Captures long-term seasonal trendsARIMA: Time series modeling for trend/seasonal decomposition
Features:
- Historical sales records
- Market conditions
- Government budget allocations
- Seasonal variations
Strategy: Just-in-Time (JIT) model to minimize holding costs
Techniques:
Linear Programmingfor efficient inventory allocation under space constraintsGenetic Algorithmsfor robust exploration of stocking strategies
Constraints:
- Maximum warehouse capacity: 5000 m³
- Forecast-aligned inventory levels
- Avoid overstocking and stockouts
- Languages: Python
- Libraries:
Pandas,NumPy,Scikit-learn,TensorFlow,XGBoost,Statsmodels - Optimization:
PuLP, customGenetic Algorithmimplementation - Visualization:
Matplotlib,Seaborn
- 🔮 Improved demand prediction accuracy across time frames
- 📉 ~20% reduction in holding costs through lean stocking
- 📦 Inventory planning remained within the 5000 m³ constraint
- 📊 Dashboards delivered actionable, real-time business insights
Our solution enables XYZ Ltd. to:
- Forecast demand with higher confidence
- Optimize inventory levels dynamically
- Make data-driven decisions using live dashboards
- Achieve cost reduction and better customer service in a volatile market
Built during SRM Datathon 2025
Team: [SubX]