Ticker trend - stock closing price prediction using machine learning

The unstable and non-linear nature of financial markets makes it complex to predict stock closing prices. The purpose of this project is to compare the effectiveness of the autoregressive integrated moving average–generalized autoregressive conditional heteroskedasticity (ARIMA-GARCH) and long short-term memory-attention residual order (LSTM-ARO) models to improve stock closing price prediction. Newly developed variables like high-low (H-L), open-close (O-C), moving averages, and standard deviations are added to past stock data in order to capture intricate patterns in stock movements. An analysis is conducted between the ARIMA-GARCH hybrid model, which successfully models linear relationships, short-term trends, and volatility clustering in time-series data, and the LSTM-ARO attention-based model, which is renowned for its capacity to capture long-term dependencies and dynamically focus on significant time steps. To evaluate the predictive accuracy of these models, the main goal of the suggested methodology is to compare their forecasting performance. Performance is typically evaluated using metrics such as the mean absolute percentage error (MAPE) and the root mean square error (RMSE).