Abstract: In order to speed up the convergence rate of the depth belief network (Deep belief NETWORK,DBN) and to improve the acquisition accuracy of the situation elements under small sample conditions, a new method based on batch normalization (Batch normalization, BN), the security situation element acquisition mechanism of the depth belief network. On the one hand, BN is added to the deep neural network to solve the problem of gradient disappearance. On the other hand, an improved active learning (improved active learning, IAL) algorithm is proposed by using the depth neural network output layer to reverse fine-tune the depth belief network, which balances the sample types by actively selecting training samples in each iteration. By selecting the training samples actively in each iteration, the algorithm can balance the sample types to improve the accuracy of the mechanism in the small sample condition. Theoretical analysis and case data simulation results show that the proposed mechanism can solve the problem of low convergence speed or gradient disappearance of the depth neural network and accurate classification of samples. The acquisition accuracy, convergence speed and algorithm complexity are better than other situation elements in this paper.