AI-DRIVEN STOCK MARKET FORECASTING USING LSTM AND ARIMA MODELS: A COMPARATIVE ANALYSIS OF FINANCIAL TIME-SERIES DATA

Abstract

The purpose of this study is to assess the accuracy of the effectiveness of the forecasting systems based on artificial Intelligence techniques with respect to financial time-series data for prediction of the stock price. The main goal of the research is to evaluate how effective traditional statistical forecast methods and deep learning models are in accurately forecasting in the turbulent financial market to identify which one outperforms the other. The study demands a special concentration on ARIMA forecasting models, LSTM forecasting models, and a proposed ARIMA-LSTM-AI integrated forecasting model to overcome the constraints of conventional forecasting models that lack the ability to describe the financial behavior in a nonlinear way, as well as the changes in financial market volatility. This study employed a quantitative research design with historical data on the stock market from financial databases, while the use of Python-based machine learning and econometric methods was employed to implement the forecasting models. The performance of the forecasting was judged by the use of RMSE, MAE, MSE, R² score and Directive accuracy. The results demonstrated that the proposed AI forecasting models show superior performance as compared with ARIMA model and LSTM model by achieving RMSE = 6.5055, MAE = 5.3025, MSE = 42.3214 and R² = 0.9881 which confirmed excellent model fitness and capability. Moreover, the model which combined Wavelet with LSTM had the highest level of directional accuracy, that is, 89.01%, indicating that the model has a better market trend predicting ability. The research will add to the body of literature on financial forecasting with artificial intelligence and have practical implications for investors, portfolio managers, and intelligent trading systems desiring more reliable forecasts and more accurate predictions of market movements.