Predicting Geopolitical Impacts on Commodity Prices
Developed LSTM forecasting models and Monte Carlo simulations at Mercedes-Benz Türk to predict Brent crude and aluminium prices.
TL;DR: Developed predictive forecasting models and risk simulations for the Technical Procurement Department at Mercedes-Benz Türk to estimate Brent crude and aluminium prices amid geopolitical conflicts, enabling data-driven contract negotiations.

Technologies & Methodology
- Forecasting & Simulation Models: Long Short-Term Memory (LSTM) neural networks, Markov Chains, and Monte Carlo Simulations.
- Statistical & EDA Tools: Shapiro test, Skewness test, Pearson correlation matrixes, cross-correlation analysis, line graphs, bar charts, and histograms.
- Data Sources: London Metal Exchange (LME), UCDP Candidate Events Dataset, WTO’s Strait of Hormuz Trade Tracker, and Federal Reserve Economic Data (FRED).
Key Contributions & Results
- Compiled, filtered, and merged disparate open-source datasets to analyze the relationship between Strait of Hormuz maritime traffic, conflict frequency, and commodity pricing.
- Extracted state transition probabilities from historical daily conflict data—categorized into main, proxy, and insurgency conflicts—using Markov Chain modeling.
- Trained two discrete LSTM models: one to predict Brent Crude prices based on historical market data and conflict states, and a second to forecast aluminium prices driven by energy costs.
- Executed 1,000 Monte Carlo simulations projecting conflict states through the end of 2026, augmenting the base simulations with LSTM predictions to map potential market volatility.
- Delivered a comprehensive risk assessment to the procurement team, establishing a 90% probability mean for Brent crude between $84.37 and $108.29, and for aluminium between $3090.83 and $3361.34.