Empirical Evaluation of Quantum Kernel Learning for Sector Rotation Prediction: Using Fama–French 10 Industry Portfolios

Quantum Machine Learning (QML) is gaining attention for extracting nonlinear struc-tures from high-dimensional data, yet its practical utility in financial forecasting remains insufficiently verified. This study empirically evaluates the effectiveness of quantum ker-nel learning for sector rotation prediction using daily returns from Fama–French 10 In-dustry Portfolios (2022–2024). We formulated a supervised classification task using a compact nine-feature set comprising mean return, volatility, win rate, excess return, MA5, MA20, max drawdown, 5-day return, and skewness. We compared classical Support Vector Classifiers (RBF and Linear kernels) against a fidelity quantum kernel (ZZFea-tureMap) evaluated on AerSimulator using 1-nearest neighbor classification. Results in-dicate that while the classical RBF kernel achieved the highest average accuracy across all sectors (52.1%), the quantum kernel (50.0%) outperformed the classical linear kernel (48.4%) and achieved the highest accuracy in the Non-Durables (55.0%) and Manufactur-ing (52.2%) sectors, demonstrating superior stability. These findings suggest that the effec-tiveness of quantum kernels in the near term is not universal but sector-dependent and conditional. Although quantum methods currently trail strong classical baselines on av-erage, future research should focus on advanced circuit design and hardware-aware evaluations to unlock their potential.