Beyond Conventional Methods: Advancing Ethereum Price Prediction through Integrated Technical, On-Chain, and Machine Learning Approaches

Ethereum's anonymity and uncontrolled cryptocurrency attraction have attracted investors. Ethereum's price dynamic inspired this study's prediction analyses. Previous study has focused on either technical analysis or on-chain analysis, leaving investors without the synergistic effects of integrating the two. This study addresses missed insights and lack of cross-comparisons by identifying variable relationships and dependencies and comparing a classical model (ARIMA), a supervised deep learning model (LSTM), and an ensemble machine learning model (XGBoost) in Ethereum price prediction. The dependent variable is Ethereum price and the independent variables are opening, high, low, closing, adjusted closing, volume traded, market capitalization, cumulative return, transactions, blocks, and gas utilized. Prices and market capitalization, traded volume, and volume are strongly correlated, and the LSTM model is the most promising due to its greater prediction accuracy and generality. The analysis reveals the bitcoin market's complexity, affecting investing and risk management.