From Data to Decisions: AI-Powered Stock Market Prediction

Abstract

Forecasting of stock markets is a complex and dynamic challenge on the account of many factors including historical trends, technical analysis and sentiment. For this research, we used Deep-learning methodologies combined with the conventional financial analysis to forecast the stock price using Long Short Term Memory (LSTM) networks. Yahoo Finance is used to download the historical stock data and basic technical indicators such as Simple Moving Averages (SMA), Moving Average Convergence Divergence (MACD), Bollinger Bands, Relative Strength Index (RSI), and Volatility to increase the predictability level. Normalization with MinMax Scaling is done to prevent the training from getting instable as part of the preprocessing of data. By feeding time series data of stock price time series, the long range dependencies and market trends are learnt by the LSTM model. Performance metrics, including Mean Squared Error (MSE), Mean Absolute Error (MAE), R² score, and the satisfaction of choosing the grade of the effect, are assessed in order to test the efficacy of the proposed model. Other suggestions for investment are derived through SMA crossovers, RSI levels, and volatility measures. The conclusion is that models based on LSTM can best mimic complex price fluctuations and make reasonable predictions than the usual statistical methods. Live stock analysis is implemented within Streamlit and shows market trends, compares current and forecast prices, and helps decision making based on intelligent insights. The contribution of this study is to demonstrate how financial time series forecasting benefits from features engineered using deep learning and also to illustrate feature engineering as an approach to improve the model resilience. Accuracies of predictions can be improved by having included Sentiment Analysis and hybrid model methodologies in future.