Abstract: The effectiveness of the genetic algorithm (GA) in the design and BPNN structure optimization and its application in the prediction of rock mass deformation modulus was researched. GA is used to find out the optimal number of neurons in the hidden layer and learning factor and momentum factor of the hidden layer and output layer，and then compared with the trial-and-error process. 76 groups of data sets derived from the actual roadway were used to validate this method. With performance criteria such as MSE，MAE and R，it is proved that GA-BPNN model is better than BPNN trial-and-error model in the prediction of rock mass deformation modulus.

Key words: Surrounding rock, Deformation modulus, Prediction, GA-BPNN