<p>Rock mass classification is a fundamental prerequisite for tunnel design, construction safety, and the optimisation of support schemes. The traditional Q-system relies heavily on engineers’ subjective interpretation of parameters such as the joint roughness coefficient and joint alteration coefficient, which often leads to classification uncertainty and errors. To improve the accuracy and objectivity of rock mass classification, a coupled evaluation method integrating fuzzy reasoning and support vector machines (SVM) is proposed. First, subjective parameters are optimised through fuzzy reasoning to reduce empirical interpretation bias. Subsequently, SVM is employed to capture the nonlinear relationships between the optimised features and rock mass classes, thereby enhancing classification performance. Finally, a standardised model training and validation framework is established to systematically verify the effectiveness of the proposed coupled model. The results indicate that the proposed method accuracy of 90%, correctly identifying all Grade V samples. For high-risk Grade IV and Grade V rock masses, both precision and recall exceed 85%, significantly outperforming the standalone SVM model. The proposed methodology effectively alleviates ambiguities arising from subjective parameter interpretation and provides a reliable and quantitative approach for tunnel rock mass classification, demonstrating strong engineering applicability.</p>