Evaluating Machine Learning Models for Short-Term Bitcoin Price Forecasting

This report evaluates the performance of supervised machine learning models for short-term Bitcoin price forecasting using historical Bitcoin and Nasdaq Composite Index data. Linear regression and random forest regression models were trained and assessed using mean squared error and R² metrics on training, validation, and test datasets. The study analyzes the predictive behavior, limitations, and apparent accuracy of the models, with particular attention to feature selection and overfitting risks in highly volatile financial time series.The results demonstrate that while the models capture short-term trends, their practical predictive value for investment decision-making remains limited.This document is a revised version of a co-authored course research project completed as part of machine learning coursework at Aalto University.