Tunnel rock quality ranks based on support vector machine

The support vector method was applied to classify rock quality, and the indexes often used in engineering such as rock quality designation, integrity coefficient, uniaxial saturated compressive strength, and friction factor of structural planes were adopted as discriminant parameters. The radial basis kernel function was selected to train samples, the optimized model parameters were determined by cross-validation, and a model of rock quality ranks was established. In comparison with the existing multi-classification model based on support vector machine constructed by a one-against-all method, the multi-classification model constructed by the pairwise method proposed in this paper may obviously reduce the indivisible region, that is, extraordinarily improves the model accuracy. Applications of this model to engineering show that the result of this model agrees with that of engineering that the classification method of rock quality ranks is effective.