Skills and Insights Gained

This project focuses on forecasting Ethereum (ETH) prices using historical data from 2016 to 2024, applying various time series models to analyze trends and volatility in cryptocurrency markets. The study involves:

- Data preprocessing, including missing value handling, anomaly detection, and transformation.
- Stationarity tests, (KPSS, ADF, HEGY) to ensure model assumptions are met.
- Model selection and forecasting using ARIMA, TBATS, Prophet, Neural Networks, and Exponential Smoothing techniques.
- Performance evaluation through MAPE, MAE, and RMSE metrics.

The results indicate that while ARIMA models provide reasonable forecasts, Neural Networks outperform traditional models in capturing complex patterns in volatile data. The project highlights the potential of machine learning in financial time series forecasting and opens avenues for further research into incorporating additional features like trading volume and market sentiment.