Table 1 The summary of the related work.
From: Enhanced stock market forecasting using dandelion optimization-driven 3D-CNN-GRU classification
Authors | Methodology | Merits | Demerits |
|---|---|---|---|
Ali et al.21 | Hybrid EMD and LSTM networks | Improved prediction accuracy using hybrid ensemble approach | Dependency on initial decomposition methods and complexity |
Zaheer et al.22 | Hybrid deep learning (CNN, LSTM, RNN) | Outperformed conventional models | Struggles with capturing complex patterns in single-layer RNN |
Brogaard and Zareei23 | Machine learning techniques, evolving genetic algorithm | Effective in generating successful trading strategies | Maintaining profitability over time due to market efficiency |
Sarma et al.24 | Analysis of stock market patterns | Comprehensive enumeration of influencing variables | May lacks robustness in real-time predictions |
Han et al.25 | N-Period Min–Max (NPMM) labeling with XGBoost | Improved trading performance | Sensitive to labeling accuracy and parameter choice |
Bhambu et al.26 | RNN, GRU, LSTM, Bi-LSTM models | Bi-LSTM outperformed others with careful hyperparameter tuning | Risk of overfitting and requires extensive tuning |
Azevedo and Hoegner27 | Extensive model comparison with machine learning algorithms | Effective in emphasizing market inefficiencies | High computational cost and complexity |
Costola et al.28 | BERT model for news sentiment analysis | Statistically significant link between news sentiment and market | Depends on quality and timeliness of news data |
Md et al.29 | MLS LSTM with Adam optimizer | High prediction accuracy | Computational overhead with very large datasets |
Gülmez30 | Optimized deep LSTM network (LSTM-ARO) | Superior performance with ARO model | Implementation complexity of hybrid models |
Shaban et al.31 | SMP-DL with data preprocessing and BiGRU, LSTM | High performance in stock price prediction | Generalizability across different market conditions |
Jarrah and Derbali32 | Multivariate LSTM-DL model | High prediction rates | Requires extensive computational resources and tuning |
Awad et al.33 | Integration of historical stock data and social media analysis | Remarkable accuracy and efficiency in predictive capacity | Dependency on social media data could introduce noise |
