SIMPLE STOCK PRICE PREDICTION USING MACHINE LEARNING AND DEEP LEARNING MODELS

Abstract

Predicting stock prices are a tough problem in financial analytics because markets are highly volatile, complex, and constantly changing. In this study, we compare traditional Machine Learning (ML) models—such as Random Forest, Support Vector Regression (SVR), and XGBoost—with a Deep Learning (DL) model called Long Short-Term Memory (LSTM) for predicting next-day stock prices. To improve accuracy, we propose a hybrid ensemble approach that combines the strength of LSTM (good at capturing time-based patterns) with tree-based and kernel-based models (known for reliable predictions). We used one year of historical stock data (Open, High, Low, Close, and Volume) collected through the Alpha Vantage API. The data was cleaned, normalized, and enhanced with lag features to better capture market behavior. We evaluated the models using error metrics such as RMSE, MAE, and R². The results show that the hybrid model performs better than any individual model. It reached an RMSE of 6.53 and an R² of 0.91, proving it to be accurate, stable, and practical for real-world stock forecasting applications.

Keywords
Stock Price Prediction, Machine Learning, Deep Learning, LSTM, XGBoost, SVR, Random Forest, Ensemble Model, Financial Forecasting, Time-Series Analysis.