A Journal on Cryptocurrency Analysis and Price Prediction Model using LSTM Neural Networks

Abstract

Cryptocurrency is a kind of virtual currency that came into existence with the recent advancement of technology in finance. It is used to complete transactions in a secure way by using the techniques of cryptography. This virtual currency is created with the help of block chain technology. In many countries, the transactions using cryptocurrency are not legalised by the banks. Some of the most popular cryptocurrencies are Bitcoin, Dogecoin, Litecoin, etc. The value of each cryptocurrency keeps varying from time to time. In this research paper, we build a data analytics model of the various cryptocurrency and also a machine learning model using the LSTM (Long Short-Term Memory) algorithm to forecast the value of a certain cryptocurrency on a particular day. This paper uses a web application called Yahoo Finance(yfinance) which has all the details of the live stock market and this acts as a source of dataset for this paper. Also, various python packages such as numpy, pandas, tensorflow, seaborn, matplotlib, etc. are used for building a model. The LSTM algorithm makes use of RNN (Recurrent Neural Network) which is powerful to model data sequencer because it has an internal memory state to store the past seen data. The proposed LSTM algorithm has a 98% accuracy which outperforms the accuracy of other existing models.