Cryptocurrency Price Prediction using Time Series and Social Sentiment Data

Abstract

With data accumulated at a rapid phase through multiple channels, algorithmic trading becomes critical in stock markets and crypto markets. In algorithmic trading, an innovative approach to integrating machine learning can provide data-driven solutions to help people invest with minimal risk and maximum returns. This study explores various machine learning techniques to model the nonlinear relationship between bitcoin prices and social sentiment data and predict the price values with some lead time. Also, the cryptocurrency market is very volatile and lacking strict governing bodies and regulators across regions making it more complex and challenging to predict the prices. Through the analysis, it is found that the sentiment data model is superior in capturing the nonlinear relationship compared to the conventional methods of technical indicators and decision trees, while the neural network models are robust and offer better accuracy in predicting bitcoin price.