Cryptocurrency Trend Analyser and Recommendation System using Deep Learning

Abstract

Cryptocurrencies are blockchain based digital currencies and have many applications in today’s world, cryptocurrencies are being accepted as a form of money in many countries. Cryptocurrencies can be traded on many exchanges and traders can benefit from trading by predicting on the right time about the possible trend of the cryptocurrency. The skill of trading requires a lot of knowledge about technical analysis and fundamental analysis. A trader should be able to time the market based on the sentiments and mood of the market and based on the prediction, an estimate of how the particular cryptocurrency can perform in the market can be formed. This project intends to make trading of cryptocurrencies easier using deep learning and machine learning techniques. First of all five cryptocurrencies,viz. Bitcoin(BTC), Litecoin(LTC), Solana(SOL), Dogecoin(DOGE) and Ethereum(ETH) are selected and datasets of all of those five are considered, downloaded from kaggle. The datasets contain historical time series data. After cleaning and preprocessing of the datasets of all the five cryptocurrency datasets, all of the datasets are split into training and testing sub datasets. Different machine learning and deep learning models are created and trained with the five sub datasets for training. After testing all the datasets on the models, a comparative analysis is performed on the models and the best model is chosen for which the models give best accuracy. In the end, a Twitter sentiment analysis is performed on the basis of various tweets of the users and the market sentiment gives a clear idea about the market analysis. On the basis of the prediction given by the deep learning model and the Twitter sentiment analysis model, the user can easily identify the trend of the market.