Bitcoin is a current popular cryptocurrency with a promising future. Itâ??s like a\nstock market with time series, the series of indexed data points. We looked at\ndifferent deep learning networks and methods of improving the accuracy, including\nmin-max normalization, Adam optimizer and windows min-max\nnormalization. We gathered data on the Bitcoin price per minute, and we\nrearranged them to reflect Bitcoin price in hours, a total of 56,832 points. We\ntook 24 hours of data as input and output the Bitcoin price of the next hour.\nWe compared the different models and found that the lack of memory means\nthat Multi-Layer Perceptron (MLP) is ill-suited for the case of predicting\nprice based on current trend. Long Short-Term Memory (LSTM) provides\nrelatively the best prediction when past memory and Gated Recurrent Network\n(GRU) is included in the model.
Loading....