Examining the relationship between environmental degradation, energy consumption, and logistics performance in OECD countries: An application using machine learning

Logistics activities contribute to climate change and global warming through greenhouse gas emissions, resulting in destructive consequences for the planet. To mitigate these effects, various measures need to be implemented both in the logistics sector and in energy production and consumption. This thesis investigates the relationship between logistics performance, environmental degradation, and energy consumption in 38 OECD countries using the Random Forest machine learning algorithm. In the analysis, MSE, MAE, RMSE, and R² metrics are used to evaluate the algorithm's performance, while TreeSHAP are employed to assess variable importance. Additionally, bee swarm plots are utilized for visualizing the results. The algorithm successfully predict the variables of energy consumption from the transportation sector, environmental degradation, 'the compentence and quality of logistics services', 'the quality of trade and transport-related infrastructure', 'the ability to track and trace consingments', and 'the efficiency of customs and border management clearance' from Logistics Performance Index (LPI) indicators; however, it has difficulty explaining 'the frequency with which shipments reach consignee within scheduled or expected delivery time' and 'the ease of arranging competitively priced international shipments' indicators. According to variable importance levels, for each target LPI indicator, other LPI indicators consistently serve as the most influential predictors. The results related to energy consumption indicate that, for accurate energy consumption predictions, priority should be given to environmental and economic factors rather than logistics metrics. The analysis also reveals that energy consumption is the most dominant determinant in forecasting environmental degradation.