"Ether Price Prediction Using Gated Recurrent Unit"

Authors

Gopinath Balu, Crescent Institute of Science and Technology, Vandalur, Chennai.

Abstract

The main objective of this project is to develop a prediction model which predicts Ether crypto currency price. Ether is the second widely used crypto currency with $60 billion market capitalization as on date. Price prediction of stocks from stock market has been well studied both in industry and academia but crypto currency price prediction is just gaining momentum. Specifically Ether price prediction is inadequately studied using Gated Recurrent Unit (GRU). GRU is a type of neural network which is mostly used for time series model and sequence models. Long Short-Term Memory (LSTM) is a predecessor of GRU which has the same accuracy but not so efficient to train. For this reason GRU architecture was used in this work and the results show the capability to handle volatile price index and remembering information and dependency from the past. Prediction from GRU model has good prediction accuracy in the test set. In this project data was scraped from a crypto currency website called CoinMarketCap and split into training set and test set. Training data was used to train the model and test data was used to check the accuracy of the model’s prediction.