Baltic dry index forecast using financial market data: Machine learning methods and SHAP explanations

Area of Science:

• Economics
• Financial Markets
• Maritime Logistics

Background:

• The Baltic Dry Index (BDI) is a key indicator of global shipping freight rates and chartering activity.
• Accurate BDI forecasting is crucial for stakeholders in the maritime and financial sectors.
• Previous research has often overlooked the impact of specific regional financial indicators on the BDI.

Purpose of the Study:

• To forecast the Baltic Dry Index (BDI) with enhanced accuracy.
• To identify and analyze the influence of diverse financial indicators, including regional ones, on BDI movements.
• To provide a deeper understanding of the economic drivers behind BDI fluctuations.

Main Methods:

• Utilized advanced machine learning algorithms: Extremely Randomized Trees, Categorical Boosting (CatBoost), and Random Forest.
• Integrated a comprehensive dataset including commodities, currencies, stock markets, and volatility indices.
• Employed the Shapley Additive Explanations (SHAP) framework for feature importance analysis.

Main Results:

• The S&P 500 index was identified as the most significant predictor of the BDI.
• Commodity indices (iron ore, coal) and the dollar index also demonstrated substantial influence.
• The study successfully integrated regional financial indicators from the U.S., EU, and Hong Kong.

Conclusions:

• The U.S. economy, reflected in the S&P 500 and dollar index, plays a pivotal role in BDI trends.
• Machine learning models, augmented by SHAP analysis, offer superior BDI forecasting capabilities.
• This research provides actionable insights for improving decision-making and stability within the global shipping industry.