STOCK PRICE PREDICTION USING MACHINE LEARNING APPROACH

Abstract

Forecasting stock market prices is a complex challenge due to the dynamic and nonlinear behavior of financial data. Accurate stock prediction is essential for assisting investors and analysts in making sound investment decisions. With recent advancements in artificial intelligence, machine learning techniques have become valuable tools for understanding market patterns and predicting future prices. These intelligent models are capable of identifying subtle correlations and hidden dependencies within large datasets that traditional statistical methods often overlook. By learning from past market trends, they can generate adaptive predictions that align more closely with real-time financial fluctuations.This study focuses on predicting and visualizing next-day stock prices using three models: Linear Regression, Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU). Historical stock data containing open, high, low, close prices, and trading volume were preprocessed to extract relevant features such as moving averages and lagged values. The models were trained and assessed using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R² score.Experimental results indicate that the Linear Regression model achieved the highest prediction accuracy of 95%, surpassing LSTM and GRU in short-term forecasting.The study demonstrates that although deep learning models like LSTM and GRU effectively capture sequential dependencies, simpler linear models can outperform them when the dataset exhibits strong linear relationships. This highlights the importance of selecting models based on the nature of market behavior and the characteristics of the data. Future research can integrate macroeconomic indicators and news sentiment analysis to further improve model performance.The performance of predictive models was analyzed using multiple evaluation metrics to ensure reliability and robustness. The study emphasizes how data preprocessing and model selection directly influence forecasting accuracy. This research contributes to understanding the efficiency of machine learning methods in short-term financial prediction and sets a foundation for future work using hybrid and real-time models.