Broad learning system (BLS) is an emerging neural network characterized by its rapid processing and robust generalization capabilities. However, determining the appropriate structure for broad learning system is also a challenge. In addition, broad learning system may perform overfitting due to the dependence between nodes in processing fully connected network. To deal with these problems, an efficient ensemble broad learning system based on Dropout and Dropconnect is proposed in this paper. The proposed Dropout ensemble broad learning system randomly discards hidden nodes to improve diversity between individuals and reduce the synergy between nodes to improve prediction stability. The Dropconnect ensemble broad learning system randomly drops connection weights to generate more complementary models by adding input attribute disturbance. The experimental results on the UCI datasets confirm that the method proposed in this paper can solve the problem of model overfitting caused by the strong dependence between the nodes of ensemble broad learning system. The proposed algorithm outperforms the original BLS in terms of prediction stability and classification accuracy.