<![CDATA[In today's financial market, precise prediction of stock prices is a daunting but critical task. This study provides a detailed description of an advanced LSTM model designed for stock price prediction where the author emphasizes the novel macro features created as well as the hyperparameter tuning performed to achieve the goal of this project. Using a dataset of 600 daily stock prices, the author performed Min-Max normalization and split the dataset into 80% training and 20% testing data for analysis. For 40 epochs, the LSTM model, which contains a single layer of 50 hidden units, was trained using the Adam optimizer and the Mean Squared Error (MSE) was defined as the loss function. The resulting Root Mean Squared Error (RMSE) of 2.85, Mean Absolute Error (MAE) of 2.35, and 0.945 of the R² values achieved in the testing dataset surpasses previously published LSTM approaches. The prediction error distributions, the scatter graphs, as well as loss training and validation plots, and most importantly the MAE and RMSE attest the model's efficient performance and convergence. The overall results support the author’s claim that the model designed provides more utility for financial forecasting. Emphasis on novel feature creation provides a more solid foundation for time series forecasting. New projects may benefit from the use of attention mechanisms, multi-source datasets, and transfer learning which will improve model predictability and generalization capability.]]>
