Developing a Hybrid Methodological Approach to Forecasting a Country’s Decarbonization Potential
Year: 2026
Type: article
Abstract: The high variability of data on decarbonization potential and the poor formalization of these processes require modern hybrid analysis methods. The study proposes an approach that combines Kohonen self-organizing maps (SOM) to identify patterns between countries, as well as linear regression (LR) and a multilayer perceptron (MLP) neural network to forecast the average CO2 emission within clusters. The analysis covers 45 countries over a ten-year period and is based on 12 explanatory indicators. Thus, the scientific methodology based on the integration of clustering and neural network forecasting to model different national decarbonization trajectories. Kohonen maps allowed authors to identify seven groups of countries, which confirms the heterogeneity of the impact of factors and patterns of decarbonization across countries. The research results showed that linear models have limited explanatory power (R² of 0.30 on test data before clustering and 0.61 in average after clustering), while the optimized neural network provides higher forecast accuracy (average R² for all clusters of 0.75 on test data) supporting clusterbased modeling and ability of neural networks to capture nonlinear relationships. The outcomes of the research can be used by policymakers to design and tailor national climate policy and plan energy transitions.
Source: JOURNAL OF EUROPEAN ECONOMY
Authors: Olena ZHYTKEVYCH, Andriy MATVIYCHUK
Language: English
Subfield: Economics and Econometrics
Domain: Social Sciences
Open Access status: diamond
OpenAlex ID: W7220409402