Paper Title
Hybrid Quantum Classical Machine Learning for Molecular Energy Estimation

Abstract
The characterization of the molecular energy properties is one of the fundamental areas of quantum chemistry for clear understanding of the stability, reactivity, and electronic attributes of the molecule. Previous work by Villalba-D´ıez proposed a hybrid Quantum Graph Neural Network–Variational Quantum Eigen solver (QGNN-VQE) framework to predict molecular energies using the QM9 Dataset for virtual screening. However, these large-molecule predictions were restricted due to limitations in Noisy Intermediate-Scale Quantum (NISQ) hardware and active space truncation. To deal with the issue, an improved hybrid QGNN model is developed here that predicts Ionization Potential (IP) and Ground State Energy (U0). These combine classical graph-based molecular representations with quantum-enhanced features to improve prediction accuracy while retaining molecular traits. The experimental results achieved an IP prediction mean absolute error (MAE) of 0.0394% and an RMSE of 0.0477%, and further yielded a U0 prediction MAE of 0.0332%, RMSE of 0.0433%, and R2 scores of 1.0000%. Keywords - Hybrid Quantum–Classical Learning, Quantum Graph Neural Networks, Variational Quantum Eigensolver, Molecular Energy Prediction, QM9 Dataset