Work

KAYA identity-based modelling of carbon emissions in West Africa using interpretable quantum-optimised support vector regression

Year: 2026
Type: article
Abstract: Accurate forecasting of carbon emissions is vital for shaping sustainable environmental policies and advancing the Sustainable Development Goals, particularly those linked to climate action. Guided by the KAYA-identity, which links emissions to population, economic activity, energy use, and technology, this study examines seven forecasting approaches for West Africa. It applies univariate methods such as the Autoregressive model (AR) and the Autoregressive Integrated Moving Average (ARIMA), alongside multivariate techniques including Multiple Linear Regression (MLR), Extreme Learning Machine (ELM), Multilayer Perceptron (MLP), Support Vector Regression (SVR), and SVR combined with Quantum Particle Swarm Optimisation (SVR-QPSO). For interpretability, Shapley values are used to assess the contribution of each variable within the models. Among all approaches, SVR-QPSO delivers the strongest predictive performance. In the training sample, SVR-QPSO lowers the forecasting error of AR by 95.81% for Côte d’Ivoire, 93.34% for Ghana, 97.05% for Nigeria, and 92.55% for Senegal. In the testing sample, it improves on AR by 96.67% for Côte d’Ivoire, 95.15% for Ghana, 97.82% for Nigeria, and 93.37% for Senegal. SVR-QPSO is identified as the leading model for Ghana and Nigeria, SVR performs best for Côte d’Ivoire, and MLP is the strongest for Senegal. The findings show that SVR-QPSO and MLP offer valuable support for future climate policy design aimed at strengthening environmental resilience.
Language: English
Open Access status: gold
Estimated APC paid: $2,690
OpenAlex ID: W7235615264
FWCI:
Cites: 47
Cited by:
Related to: 10