As the next generation of radio astronomical telescopes continue to improve and develop, the pulsar survey will produce millions of pulsar candidates, which pose considerable challenges for pulsar identification and classification. The rapidly evolving artificial intelligence (AI) techniques are being used for pulsar identification and discovery of new pulsars. Using the pulsar data set observed with the Parkes telescope, the High Time Resolution Universe Survey (HTRUS), a 14-layer deep residual network was designed (called Residual Network, ResNet) for pulsar candidate classification. In the HTRUS sample, the number of non-pulsar candidates (i.e., negative sample) is much larger than that of pulsar candidates (i.e., positive sample). The imbalance of positive and negative samples is prone to result in model misinterpretation. By using the oversampling technique to enhance the data of positive samples in the training set and adjust the ratio of positive and negative samples, we solve this Non-equilibrium problem. During training, the hyperparameters were adjusted using 5-fold cross-validation to construct the model. The test results show that the model can achieve high precision (98%) and recall (100%), respectively, and the F1-score is able to reach 99% for each sample tested. It takes only 7 ms to complete each candidate classification, providing a viable approach to future pulsar big data analysis.