Cancer is one of the most dangerous threats to human health. One of the issues is that drug re-sistance action which leads to the side effects following drug treatment. Numerous therapies have endeavored to relieve the drug resistance action. Recently, anticancer peptides would be a novel and promising anticancer candidate for the therapy of cancers which can inhibit tumor cell pro-liferation, migration, suppress the formation of tumor blood vessels with less side effects of drug resistance. However, it is costing, laborious and time consuming to identify anticancer peptides by biology experiment through high throughput way. Therefore, how to accurately identify an-ti-cancer peptides become a key and indispensable step for anticancer peptides therapy. Although some existing silicic methods have been developed to predict anticancer peptides, the accuracy still needs to improve. Thus, in this study, we propose a deep learning-based model, called AC-PNet, to distinguish anticancer peptides from non-anticancer peptides (non-ACPs). ACPNet em-ploys three different types of features which are peptide sequence information, peptide physico-chemical properties and auto-encoding features linking the training process. ACPNet is a hybrid deep learning network, which fuses fully connected networks and recurrent neural networks. The comparison with other existed methods on ACPs82 datasets shows that ACPNet not only achieves the improvement of 1.23% Accuracy, 2.04% F1-score, 7.3% Recall but also gets balanced performance on multiple matrices, such as matthews correlation coefficient. Meanwhile, ACPNet is verified on an independent dataset with 10 proven anticancer peptides, and only one anti-cancer peptide is predicted as non-ACPs. The comparison and independent validation experiment indicate that ACPNet can accurately distinguish anticancer peptides from non-ACPs. The source codes and datasets are available at https://github.com/abcair/ACPNet.