Review of Machine Learning Techniques for Cryptocurrency Price Prediction

EasyChair Preprint 10190

Abstract

Cryptocurrency is a class of digital asset that is very challenging to monitor and forecast. Predicting cryptocurrency price action and its locus is difficult because it does not coincide with market movements. Our objective is to analyze the machine learning algorithms used in 9 researches and find out the best model which can be used to forecast the prices of time series models. In this work, we compared and analyzed earlier methodologies in which several machine learning models were applied to forecast the trend of cryptocurrency time series data. The outcomes support the machine learning models' ability to predict trends reasonably well. Making long-term predictions and generalizing them based on a small number of models yields low accuracy outputs for a highly volatile asset like cryptocurrency. We suggest using various models to fill this gap. With this method, we'll strive to identify the ideal machine learning algorithm for achieving the best accuracy with the lowest possible error rates.

Keyphrases: ARIMA, Cryptocurrency, LSTM, Prophet, XGBoost, machine learning, time series

@booklet{EasyChair:10190,
  author    = {Shubham Bhattad and Stefin Sunnymon and Dallas Vaz and Chhaya Dhavale},
  title     = {Review of Machine Learning Techniques for Cryptocurrency Price Prediction},
  howpublished = {EasyChair Preprint 10190},
  year      = {EasyChair, 2023}}