Comparative Evaluation of Statistical Models and Machine Learning Approaches in Modelling the Energy Dependency of the BIST Industrial Index: Balancing Predictive Performance and Interpretability
Comparative Evaluation of Statistical Models and Machine Learning Approaches in Modelling the Energy Dependency of the BIST Industrial Index: Balancing Predictive Performance and Interpretability

Summary/Abstract: Energy dependency plays a pivotal role in shaping the performance of stock markets, particularly in energy-sensitive indices such as the BIST Industrial Index in Turkey. This study presents a comparative evaluation of traditional statistical models and machine learning (ML) techniques in capturing the complex relationship between energy variables and the BIST Industrial Index. A dataset encompassing energy imports, production levels, and energy prices is utilised to assess the effectiveness of Ordinary Least Squares (OLS) regression, Random Forest (RF), and Gradient Boosting (GB) models. The results reveal that ML models substantially outperform traditional statistical methods in their ability to capture nonlinear, intricate relationships between energy metrics and market behaviour. Among the ML models, RF demonstrates the highest predictive accuracy. Feature importance analysis identifies crude oil production as the most significant variable, underscoring the dominant influence of domestic energy dynamics in shaping the BIST Industrial Index. While ML models offer superior forecasting capabilities, they introduce challenges in terms of model interpretability. In contexts where transparency is crucial, statistical models such as OLS remain more favoured for their simplicity and explainability. The findings highlight the need for a balanced approach in model selection, with hybrid models potentially offering the best of both worlds by combining the strengths of traditional and modern methodologies. The insights derived from this study can inform policymakers and investors, particularly within emerging markets, providing a nuanced understanding of the trade-offs between predictive power and model transparency in forecasting energy-sensitive financial indices.

• Details
• Contents