Predicting stock market movements is a complex and challenging task due to many factors that can influence stock prices. While there are various approaches and techniques that researchers and analysts have explored, it's crucial to note that accurately forecasting stock prices consistently is complicated. Financial data can be noisy and subject to errors. Minor inaccuracies in data can lead to inaccurate predictions. In this research, we propose a new deep learning technique with an optimization algorithm to overcome the limitations in stock market prediction. This research has five stages to predict closing prices using historical data. The input data is initially collected from three datasets (Apple,Microsoft,&Tesla). This data must undergo punctuation removal, outlier removal, replacing missing values, data normalization, and data cleaning performed in the pre-processing stage for accurate prediction. After that, the features are extracted using the Improved ResNeXt approach to improve the prediction performance. Finally,a novel DL technique of EfficientNet V2 is employed to predict the future stock closing prices based on historical data. To improve the prediction accuracy, the study utilizes the Multi-Strategies Improved Beluga Whale Optimization Algorithm (MSIBWOA) to optimize the hyper-parameters. For determining the results of the DL stock forecasting approaches, the Accuracy, Mean Squared Error (MSE), Pearson's Correlation (R), Normalized Root Mean Squared Error (NRMSE), and Root Mean Squared Error (RMSE) metrics were used. The performance of the proposed method is compared with other ``state-of-the-art'' techniques. In comparison to the most advanced baseline algorithms, the evaluations demonstrate a significant improvement in the prediction's performance.
Reddy et al. (Tue,) studied this question.