Stock Market Trends and Forecasting: A Study of Machine Learning and Time Series Models

The stock market plays a crucial role in economic growth by enabling efficient capital allocation and investment. This study focuses on forecasting stock closing prices and identifying future trends using a combination of machine learning and time series models. To enhance predictive accuracy, ensemble learning is employed, integrating classifiers for real-time forecasting. Real-time stock price data from Yahoo Finance is used, applying models such as Prophet, ARIMA, regression, decision trees, and long short-term memory (LSTM) networks. LSTM performs well in short-term forecasting, while decision trees excel in long-term predictions. Performance evaluation is conducted using root mean square error (RMSE) and R-squared (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>). Experimental results on Apple, Google, and Microsoft stock datasets demonstrate varying model performances across different time horizons. LSTM shows high accuracy for short-term predictions, achieving an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of 0.82 for Google stock over five days. Decision trees perform well for long-term forecasts, with R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> reaching 0.89 for Microsoft stock. Prophet provides robust trend detection, though with some deviation from actual values, while ARIMA exhibits moderate performance across different durations. RMSE results highlight that lower values correspond to more accurate models. The study confirms that combining multiple models enhances predictive reliability, offering valuable insights for stock market forecasting.