<p>Classification is fundamental in machine learning (ML) applications such as medical diagnosis, financial analysis, and image recognition. Quantum machine learning shows promise in enhancing feature representation and computational efficiency. However, challenges like the noisy intermediate-scale quantum hardware limitations, lack of interpretability, and barren plateau issue in parameterized quantum circuits hinder its practical adoption. To address these issues, this work proposes a novel hybrid approach called quantum-enhanced kernel logistic regression (QEKLR) for classification tasks, which combines quantum computation with classical logistic regression (LR) model for improved classification performance. In our QEKLR model, classical data (CD) are first transformed to quantum states (QS) using shallow-depth ZZFeatureMap, and then, fidelity kernel is employed to compute the dot product between feature vectors, resulting in a trained quantum kernel. Subsequently to predict new instances, the quantum kernel is integrated into a classical LR model. We evaluated the proposed model’s effectiveness using four datasets: synthetic, Iris, Statlog Heart Disease (HD), and Ecoli datasets achieving accuracy rates of 100%, 100%, 94.87%, and 71%, respectively. In addition, experimental results reveal that the proposed QEKLR model surpasses classical ML models, achieving the MCC 0.88 and AUC-ROC 0.98 for the Statlog HD dataset demonstrating strong classification capability. However, on the Ecoli dataset, the model’s performance was comparable to existing ML models, indicating balanced performance across datasets while maintaining robust classification ability and effective class distinctions.</p>