Cryptocurrency Market Volatility and Financial Risk Management: A Machine Learning-Based Framework for Investment Decision-Making

The nature of cryptocurrency markets is highly volatile, volatile in price and has a complex risk profile which makes predictive forecasting and investment decision making a real challenge. This paper proposes an integrated framework that integrates the areas of volatility analysis, financial risk assessment, explainable artificial intelligence, and portfolio decision support using machine learning. Daily observations of five cryptocurrencies—Bitcoin, Ethereum, Binance Coin, Cardano, and Solana—from August 2020 to September 2026 were analyzed. Next-day absolute-return forecasting models, including historical baseline, random forest (RF), XGBoost, GARCH(1,1), and LSTM, were tested. The financial risk was evaluated by Value at Risk (VaR), Expected Shortfall (ES) and statistical back testing. The test MAE (0.015123) and RMSE (0.022192) for the historical baseline were the lowest, better than advanced forecasting methods. The gross cumulative returns were the highest for XGBoost allocation (−8.787%), and inverse-volatility allocation (−9.433%) was best when transaction costs were 10 bps. The risk-aware strategies both showed a lower volatility portfolio compared to equal weighting. SHAP analysis revealed the historical volatility indicators as important predictors. The results suggest that complexity in forecasting does not necessarily equate to improved investment performance and highlight the need for interpretable and statistically sound, yet cost-effective financial decision making.