The aim of this dissertation is to investigate the use of Machine Learning (ML) methods for the prediction of Cryptocurrency price trajectories. Using a combination of price & supply/demand data (from exchanges) and sentiment (from news articles and blogs) the performance of a range of Machine Learning methods, of progressive complexity, will be evaluated. The cryptocurrency of choice will be Cardano(ADA) and the ML methods that will be used for forecasting are the following: OLS regression (including lasso regression and ridge regression), Random forests (i.e. tree based ), Support Vector Machines (SVM) regression and, time permitting, feed-forward neural network, LSTM neural, Gaussian process regression. Computations will be performed using the R-programming language (in Rstudio) and pre-built packages implementing the aforementioned algorithms. The prediction problem will be formulated both as a regression problem (i.e. try to predict the price) and as a classification problem (try to predict the direction of future price moves). The input data will come from cryptocurrency exchanges and new sources/blogs that are in the open domain.
Structure:
Introduction
Review of the literature
Methods and data
Results
Discussion and conclusions
