A Data-Driven Evaluation Method for Tunnel Rock Stability Based on Multi-Core Support Vector Regression

With the increasing complexity and scale of tunnel engineering construction, accurately assessing the stability of tunnel rocks has become particularly important. This work proposes a data-driven Multi-Kernel Support Vector Regression (MK-SVR) model for evaluating the stability of tunnel rocks. First, after considering the unique properties of the kernel function, the data characteristics of the variation sequence of tunnel rock stability parameters, and the complexity of algorithm implementation, this paper selects the polynomial kernel function and Gaussian radial basis kernel function as the basis for constructing a multi-kernel support vector regression model. Then, by weighting and combining into a composite kernel, an MK-SVR model is established to enable the model to more flexibly adapt to the nonlinear and multimodal characteristics of the data. Finally, the adaptive Artificial Fish Swarm Algorithm (AFSOA) was used to optimize the kernel parameters and improve the prediction accuracy of the model, resulting in the construction of the MK-SVR-AFSOA model for accurate evaluation of tunnel rock stability in this paper. The results indicate that the MK-SVR-AFSOA model has high accuracy and robustness in tunnel rock stability evaluation, and can effectively reduce errors and uncertainties in traditional methods.