A Systematic Literature Review For Stock Price Prediction with Machine Learning

This systematic literature review evaluates stock price prediction methods, focusing on studies that use machine learning, time series analysis, and statistical models. After applying strict inclusion and exclusion criteria, we analyze studies based on their objectives, methodologies, datasets, and results. The systematic literature review findings of these 10 studies reveal important insights and implications in addressing these 3 research questions. The review emphasizes the success of AI models, particularly in stock market forecasting, Long Short-Term Memory (LSTM) networks are used because of their capacity to analyze sequential data and identify long-term dependencies. Sentiment analysis from sources like intermarket data, news, and social media are used to further enhances predictive accuracy, especially when combined with traditional technical analysis. Market factors such as past returns, momentum, implied volatility (VIX), and cash flow financial data improve prediction precision by providing additional context. Common evaluation metrics, including accuracy, ROI, profitability, and Mean Absolute Error (MAE), demonstrate that AI models outperform traditional methods. However, challenges remain, such as reliance on noisy sentiment data, overfitting risks, and potential biases that comes from unstructured sentiment data and the relatively small amount of sentiment data. Future research could improve sentiment analysis by adding more sentiment data and quality control, integrate multiple AI techniques, and include broader financial and economic data to enhance model robustness. This review aims to highlight the increasing potential of deep learning to refine stock market prediction strategies.