Due to the trend of smart contracts and digital banking, more and more people pay attention to cryptocurrencies in recent years, and Bitcoin and Ethereum are currently the most concerned projects. Therefore, there are more and more researches on deep learning technology that apply the technology originally applicable to the stock market to this problem. Among them, the effective one is the Recurrent Neural Network (RNN) used to solve the prediction of time series problem. However, as time goes by, the vanishing gradient problem will be generated. Therefore, a Long Short-Term Memory (LSTM) model improved by RNN is proposed to solve vanishing gradient problem, but it still cannot completely solve this problem. In addition, since the price of Bitcoin is more volatile compared with the stock market, it is difficult to make accurate predictions by using traditional LSTM. Therefore, this paper proposes a new “Bi-Long Short-Term Memory based on News Technical indicators, Bi-NTLSTM” to combines news technical analysis indicators to solve improve the model’s learning ability.