Abstract: In order to predict the stability classification of roadway surrounding rock accurately and rapidly, a prediction model for stability classification of roadway surrounding rock based on support vector machine (SVM) optimized by grid search method (GSM) was proposed. Twenty-two groups of roadway surrounding rock data were chosen as learning samples, four indexes including angle between the horizontal crustal stress and roadway, roof lithologic character, water effect and basal area of roadway were chosen as model input, and roadway surrounding rock stability was chosen as model output. Meanwhile, in order to improve generalization and prediction precision of the model, the SVM parameters were optimized by improved grid search method, finally the roadway surrounding rock prediction model based on GSM-SVM was established. Then using this model to predict eight groups of roadway surrounding rock data, the prediction results were compared with those predicted by BP neural network model. The results show that the prediction results of GSM-SVM model accord with the actual results, and the correct rate reaches to 98%, so GSM-SVM model has higher accuracy than BP neural network model.