In recent years, deep learning methods have successfully solved many fake news problems. Investing in fake news detection is an important for modern people's life, notably manager. In this paper, we utilized multiple factors for detecting and fighting fake news. It is treated as time series problem. The state-of-the-art natural language processing tool BERT is used to recognize the context of text, and the Hybrid of Deep CNN-Bi-Directional LSTM model, convenient in temporal dimension analysis, is applied to detect fabricated information. We have incorporated context as an important feature to improve the accuracy: we investigates the proposed method performance using two different data sets, showing, with the results that the proposed model, namely context-based a hybrid of Deep CNN-BLSTM approach performs well, in comparison with other results and methods.