Bitcoin, one of the major cryptocurrencies, presents great opportunities and\nchallenges with its tremendous potential returns accompanying high risks. The\nhigh volatility of Bitcoin and the complex factors affecting them make the\nstudy of effective price forecasting methods of great practical importance to\nfinancial investors and researchers worldwide. In this paper, we propose a\nnovel approach called MRC-LSTM, which combines a Multi-scale Residual\nConvolutional neural network (MRC) and a Long Short-Term Memory (LSTM) to\nimplement Bitcoin closing price prediction. Specifically, the Multi-scale\nresidual module is based on one-dimensional convolution, which is not only\ncapable of adaptive detecting features of different time scales in multivariate\ntime series, but also enables the fusion of these features. LSTM has the\nability to learn long-term dependencies in series, which is widely used in\nfinancial time series forecasting. By mixing these two methods, the model is\nable to obtain highly expressive features and efficiently learn trends and\ninteractions of multivariate time series. In the study, the impact of external\nfactors such as macroeconomic variables and investor attention on the Bitcoin\nprice is considered in addition to the trading information of the Bitcoin\nmarket. We performed experiments to predict the daily closing price of Bitcoin\n(USD), and the experimental results show that MRC-LSTM significantly\noutperforms a variety of other network structures. Furthermore, we conduct\nadditional experiments on two other cryptocurrencies, Ethereum and Litecoin, to\nfurther confirm the effectiveness of the MRC-LSTM in short-term forecasting for\nmultivariate time series of cryptocurrencies.\n