Abstract
Recently, cryptocurrency investment is one of the most interesting investment channels offered on the market. Investors are always trying to find an effective prediction model to increase profit as well as reduce loss in investment. In this paper, a combination of technical indicators and deep learning is applied to predict cryptocurrency price trends in the short term. The work also used the Multi-scale Residual Convolutional (MRC) module for feature extraction and a Long Short-Term Memory (LSTM) to predict price trends. Experimental results using Bitcoin and Ethereum time-series data in time frame as 1-hour and 30-minute show that our proposed method has better accuracy in a comparison to some other methods. Moreover, an automation trading bot using the proposed model was tested on the Binance Sandbox environments and got good results.