EXPLORATORY ANALYSIS OF STOCK MARKET PREDICTION USING AUTOREGRESSIVE AND MACHINE LEARNING ALGORITHMS

With globalization and the increasing popularity of financial markets, a growing number of individuals and institutions are participating in stock investment. Predicting stock prices holds various level of strategic importance for participants in the financial markets. This study evaluates the efficacy of two methodologies, the Machine Learning (ML) algorithm – Long Short-Term Memory (LSTM) and Time Series Analysis (TSA) – Autoregressive Integrated Moving Average (ARIMA), in forecasting the Standard & Poor’s 500 (S&P 500) Index prices and provides a comprehensive comparison of their accuracy. With an explosion of financial market data, accurate stock price forecasts are essential for investment decisions. Nevertheless, the significant volatility and non-linear attributes of financial time series make it challenging for a singular method to completely encapsulate their patterns. This study employs the LSTM model as non-linear modelling instrument with ML, and ARIMA serves as benchmark method for traditional statistical model. Both approaches utilized standardized parameter optimization and training strategies, applying data collected from 2013 to 2024 to predict stock prices. In the initial period, 90% of the data (from 2013 to 2023) is designated as training data, while 10% (from 2024) is allocated for testing in the subsequent period. Following the model predictions, the Mean Squared Error (MSE) and Root Mean Squared Error (RMSE) are employed to assess the accuracy of the LSTM and ARIMA models and determine their appropriateness for the Index. The findings indicate that the LSTM model possesses considerable advantages in capturing non-linear patterns and long-term dependencies, whereas the ARIMA model exceeds in fitting short term linear trends, albeit with limited capacity to manage non-linear structure and abrupt fluctuations.