AI-Based Stock Risk Classification and Portfolio Analysis

Stock market investment is characterized by uncertainty arising from price volatility, market conditions, company performance, economic events, and changing investor behaviour. Traditional stock applications generally emphasize price and profit/loss information, while risk at the individual-stock and portfolio levels may receive less direct attention. This review examines the approaches presented in the selected research papers and relates them to the proposed problem of AI-Based Stock Risk Classification and Portfolio Analysis. The reviewed works cover deep learning for stock selection and prediction, machine-learning-based covariance estimation, Conditional Value-at-Risk (CVaR) optimization with direct reinforcement learning, LSTM-based portfolio platforms, and combinations of recurrent neural networks with genetic algorithms. The literature indicates that AI models can improve prediction, adaptive risk estimation, and portfolio decision support, but the reviewed studies generally address different parts of the investment pipeline rather than a unified stock-level Low/Medium/High risk classification followed by portfolio-level risk analysis. Based on this gap, an integrated framework is identified in which historical stock data are preprocessed, financial risk indicators such as returns, volatility, moving averages, RSI, beta, and maximum drawdown are derived, machine learning is used to classify stocks into risk categories, and portfolio concentration, diversification, and overall risk exposure are subsequently analyzed. The review also highlights important limitations, including dependence on historical data, market-regime changes, transaction costs, model generalization, and the difficulty of defining reliable risk labels.