Open Access

ARTICLE

A Quantum-Assisted Hybrid Learning Framework for Environmental CO₂ Emission Analysis

Merve Sinem Karahan^*, Mehmet Karaköse

Department of Computer Engineering, Faculty of Engineering, Firat University, Elazıg, Türkiye

* Corresponding Author: Merve Sinem Karahan. Email:

Journal of Quantum Computing 2026, 8, 101-121. https://doi.org/10.32604/jqc.2026.078969

Received 12 January 2026; Accepted 23 June 2026; Issue published 21 August 2026

Abstract

Accurate prediction of carbon emissions is essential for developing sustainable environmental policies and mitigating global warming. Road transportation represents one of the major sources of global CO₂ emissions due to its dependence on fossil fuels. This study presents a comparative framework that evaluates classical machine learning models alongside a hybrid quantum–classical learning architecture for vehicle-based CO₂ emission prediction. A large-scale vehicle emissions dataset containing 7385 samples collected over approximately seven years was obtained from the official open-data platform of the Government of Canada. Key vehicle characteristics, including engine size, fuel consumption, transmission type, and vehicle class, were utilized as predictive features. Data preprocessing involved the removal of incomplete records and normalization of numerical variables using MinMaxScaler. Three classical machine learning models, namely Random Forest, Gradient Boosting, and Multi-Layer Perceptron (MLP), were trained and evaluated using identical preprocessing procedures and evaluation protocols. Performance was assessed using the coefficient of determination (R²) and Mean Absolute Error (MAE). Among the classical approaches, the MLP achieved the highest predictive performance with an R² score of 0.963 and a normalized MAE of 0.082, followed by Gradient Boosting (R² = 0.923, MAE = 0.085) and Random Forest (R² = 0.890, MAE = 0.091). Subsequently, a hybrid quantum–classical framework integrating a variational quantum circuit with a classical neural network was developed. The proposed model achieved a normalized MAE of 0.048 and an R² score of 0.9925, outperforming all evaluated classical models. Additional experiments investigating entanglement depth demonstrated that stronger entanglement structures contributed positively to predictive performance, highlighting the importance of quantum-enhanced feature representations. Overall, the findings indicate that hybrid quantum–classical learning can provide improved predictive accuracy for structured environmental datasets. While the present study does not claim a formal quantum advantage in terms of computational complexity or scalability, the observed empirical improvements suggest that quantum-enhanced feature representations can serve as a promising complementary mechanism for environmental data analytics and sustainability-oriented decision support systems.

---

Keywords

---