As cryptocurrencies redefine global finance, accurate trend forecasting becomes essential for data-driven decision-making. This work proposes an approach to forecasting trends in the Bitcoin market using on-chain data and supervised machine learning. Instead of predicting exact price values, the focus is on identifying market direction, which may benefit not only investors, but also data scientists and risk analysts involved in financial forecasting. Data preprocessing included labeling with moving averages, normalization, and dimensionality reduction using t-SNE. Models were evaluated using TimeSeriesSplit, with precision as the primary metric. The results showed consistent performance across the classifiers, with average precision ranging from 65% to 70%, highlighting the potential of trend-based forecasting in highly volatile markets.