Open research questions in Stock Market Forecasting Methods

What the literature leaves open

• This study proposed an optimized BWOA–ACO–GRU–MLP framework for house price prediction using structural, locational, and amenity-related housing attributes. The main implication of the proposed framework is that combining wrapper-based feature selection with hyperparameter optimization can improve the stability and predictive behavior of neural regression models for real estate valuation. In particular, BWOA helped identify a compact subset of relevant predictors, while ACO reduced manual hyperparameter selection and improved model convergence. The GRU component was used as a gated feature-interaction encoder rather than as a temporal sequence model, and the MLP component performed the final nonlinear regression mapping. The experimental findings indicate that the proposed framework achieved promising predictive performance on the evaluated dataset compared with the baseline models. However, these results should be interpreted carefully. The dataset contains only 500 housing records, and the original holdout test set includes only 50 samples. Therefore, the reported performance may be sensitive to the data split and may not fully represent the variability of larger real estate markets. Although cross-validation was added to reduce this concern, external validation remains necessary before making strong claims about real-world generalizability. Another important limitation is that the reported error metrics were calculated on the normalized target scale. Therefore, MSE, RMSE, MAE, MAPE, and MedAE should be interpreted as normalized-scale comparison metrics rather than direct monetary errors in the original house-price unit. Future work should report both normalized-scale and inverse-transformed price-scale errors to improve practical interpretability for real estate decision-making. In addition, although the GRU layer improved performance in this study, house price data are tabular rather than naturally sequential. Thus, the GRU should be understood as a gated nonlinear representation-learning component, and further ablation studies are needed to confirm its contribution compared with simpler tabular models. Future research should validate the proposed framework on larger, geographically diverse, and temporally updated housing datasets. Additional comparisons with strong tabular machine-learning models such as XGBoost, CatBoost, LightGBM, Random Forest, and SVR should also be maintained to ensure fair benchmarking. Furthermore, explainability methods such as SHAP, permutation importance, or partial dependence analysis should be incorporated to clarify how selected features influence predicted prices. These future improvements would strengthen the practical reliability, interpretability, and generalizability of the proposed framework for intelligent real estate valuation systems.

  House price prediction using a hybrid GRU–MLP based on binary whale optimization algorithm and ant colony optimization for hyperparameter tuning · 2026 · DOI

• In essence, we compared using 10-year rolling windows five single models (ARIMA, Logistic Regression, Random Forest, XGBoost and LSTM) with a metamodel designed to predict the sign using two rules – majority, if four out of five models predict the same sign, or a fallback mechanism if there is no consensus among individual models.

  The Effectiveness of Metamodeling in Predicting the Sign of Asset Returns: An Empirical Assessment · 2026 · DOI

• The Real-Time Sentiment-Driven Stock Market Simulator demonstrates how user-generated social media content can influence simulated stock price movement in a controlled environment. The system connects tweet creation, sentiment analysis, impact calculation, stock simulation, rule-based prediction, database storage, and visual output into a complete full-stack web application. The project provides a useful educational platform for understanding the relationship between sentiment and stock price behavior without depending on actual stock market data or real trading activity. By comparing actual simulated price and predicted price using Mean Absolute Error, the system allows users to observe the behavior and limitations of a rule-based prediction engine inside the simulator. Future work can improve the system by replacing the rule-based prediction engine with a trained forecasting model such as ARIMA or LSTM. The sentiment analysis module can also be integrated directly into the main application. Additional factors such as user influence, follower count, company news, real market APIs, technical indicators, and backtesting can be added to make the simulator more realistic and analytically useful. © Author(s). This work is peer-reviewed, openly published, and permanently archived This article is openly accessible and reusable with proper attribution.

  Stock Market Prediction and Sentiment Analysis · 2026 · DOI

• Figure 2. Research Framework 3.2 Data Collection and Preparation The dataset in Table 2 consists of 4,128 daily observations of ANTM.JK stock data from September 29, 2005, to May 31, 2022. The data were chronologically divided into 2,683 training observations (65%) and 1,445 test observations (35%). Only the Close Price variable was used as the forecasting target. The Close Price was selected because it represents the final market valuation at the end of each trading session and is the most used variable in univariate stock forecasting studies. Although the dataset covers a long historical period, the COVID-19 and Omicron periods are used as the main volatility context because ANTM.JK experienced substantial price fluctuations during this period. Table 2.

  A Robustness-Oriented Evaluation of LSTM, GRU, and Hybrid LSTM-GRU Models for ANTM.JK Stock Price Forecasting · 2026 · DOI

• This paper successfully designed, implemented, and validated an adaptive, data-driven machine learning Abd‑Elwahab et al. Future Business Journal (2026) 12:191 Page 13 of 14 framework for forecasting Telecom Egypt’s stock price. Its primary contribution lies in providing empirical evidence that deep learning architectures, specifi- cally LSTM, achieve optimal and reliable forecasting performance when leveraging fundamental financial ratios within an emerging market context. The study demonstrates that a carefully constructed pipeline inte- grating financial theory with sequential modeling can significantly enhance predictive accuracy over conven- tional single-perspective approaches. Furthermore, the development and comparative analysis of the frame- work using technical, fundamental, and hybrid datasets offer a valuable, structured methodology for financial analysts and investors. The findings ultimately highlight that the choice of predictive inputs aligned with the model’s strengths is paramount, delivering a practical tool for improved investment decision-making in envi- ronments like the Egyptian stock exchange. Future work should explore advanced feature selec- tion, attention-based models, and multi-asset data- sets to improve hybrid modelling performance and generalizability. Building upon the findings and limitations of this study, the following directions are proposed for future research: • Integrating attention-based deep learning models or transformer architectures. • Using ensemble methods to combine model strengths and reduce overfitting. • Applying explainable AI techniques to increase model interpretability. • Extending the framework to include multiple stocks, indices, or even multi-market predictions.

  A hybrid machine learning framework integrating technical and fundamental analysis for stock market prediction · 2026 · DOI

• The results indicate that four years of single-instrument five-minute OHLCV data are insufficient for reliable sequential ML-based intraday forecasting.

  Sequential Structure in Intraday Futures Data: LSTM vs Gradient Boosting on MNQ · 2026

• This study provides a comprehensive evaluation of bagging ensemble models for financial time series (FTS) classification and addresses a gap in the literature regarding how bootstrap methods, ensemble sizes, voting mechanisms, and loss functions jointly influence model performance.

  Ensemble Approach for Financial Time Series Modeling · 2026 · DOI

• Within this set, the autoregressive benchmark is restricted to an AR(1) specification; more flexible but still par- simonious time-series models such as AR(p), ARIMA, HAR-RV or GARCH-type frameworks are not explored.

  Forecasting cryptocurrencies in turbulent times: Evidence on parsimony versus model complexity · 2026 · DOI

• The current implementation acknowledges the following limitations: • External Dependency: System performance is contingent on the availability and reliability of the Yahoo Finance API. Any downtime or rate limiting on Yahoo Finance's end directly impacts data retrieval. • Prediction Accuracy: While Facebook Prophet is well-suited for trend forecasting, it does not incorporate real-time news, earnings announcements, or sentiment signals, which are significant drivers of short-term stock price movements. • Internet Dependency: The application requires an active internet connection for all operations; no offline mode is currently supported. © 2026 The Author(s). Published by IJCOPE Journal. Website: https://ijcope.org/ 6 International Journal of Creative and Open Research in Engineering and Management ISSN: 3108-1754 (Online) Volume 02 Issue 05 May-2026 | Impact Factor: 3.5 • Stock Universe: The system currently supports only stocks available through the Yahoo Finance API. Exotic instruments, unlisted securities, and some regional exchanges may not be accessible. • Single-Day Forecast: The primary prediction output is the next-day price; multi-step probabilistic forecasts with uncertainty visualization are not yet exposed in the frontend. 9. FUTURE SCOPE Several enhancements are planned to extend the capabilities of the system: • Sentiment Analysis Integration: Incorporating real-time news and social media sentiment (e.g., from Twitter/X and financial news APIs) as auxiliary features to improve short-term prediction accuracy. • Advanced ML Models: Exploring LSTM and Transformer-based architectures to complement or replace Prophet for stocks with irregular seasonality patterns. • Portfolio Tracking: Allowing users to create and monitor a personal portfolio, with aggregate risk and return metrics displayed on a dedicated dashboard. • Mobile Application: Development of a native mobile application (Android and iOS) to improve accessibility and enable push notifications for price alerts. • User Authentication: Implementing secure user accounts with persistent watchlists and prediction history. • Multi-Step Forecasting: Exposing 7-day and 30-day probabilistic forecasts with confidence band visualizations to support longer-horizon investment planning. 10. CONCLUSION This paper presented the design, architecture, and implementation of an AI-Powered Stock Analysis and Prediction Application that bridges the gap between raw financial market data and actionable investment insights.

  AI-Powered Stock Analysis and Prediction Application · 2026 · DOI

• The model's capacity to handle unforeseen circumstances or abrupt market shocks is limited by its dependence on historical data. Furthermore, there may be difficulties due to the computational complexity, particularly when scaling the model for real-time applications or enormous datasets. Our experiments primarily focused on four sectors: Banking, Automobile, Steel, and Pharma. While these sectors provide a range of market behaviors and volatility levels, limiting the evaluation to a specific subset may introduce sector- specific biases. As a result, the generalizability of the model to other sectors (e.g., technology, energy, consumer goods) remains to be validated. Expanding the sectoral diversity of the dataset in future work will help better assess the model's robustness across broader market conditions.

  Hybrid TCN-transformer Model with Multi-head Attention for Stock Price Forecasting · 2026 · DOI

• While this review provides a comprehensive synthesis of machine learning applications in stock price prediction, several limitations should be acknowledged (Snyder, 2019).

  A SYSTEMATIC LITERATURE REVIEW AND BIBLIOMETRIC ANALYSIS OF MACHINE LEARNING ALGORITHMS AND TECHNICAL INDICATORS TO STOCK PRICE PREDICTION · 2026 · DOI

• Implications for Research Academically, the growing popularity of MLbased forecasting systems opens the prospects of SCIENTIFIC CULTURE, Vol. 12, No. 4, (2026), pp. 9514-9527 9524 A SYSTEMATIC LITERATURE REVIEW AND BIBLIOMETRIC ANALYSIS methodological development and more robust theoretical assimilation (Henrique et al., 2019). Despite the consistency of the results of higher predictive accuracy provided by the state-of-theart architectures, including Long Short-Term Memory (LSTM) networks, a significant portion of studies is still research of a rather accuracyoriented and insufficiently supported by financial theory (Makridakis et al., 2018). Future studies ought to go beyond the idea of progressive error minimization and focus on robustness testing in different market regimes, structural breaks, and different asset classes (Lopez de Prado, 2018). The hybrid models combining the methods of ML with the conventional econometric models (e.g., ARIMA) show a significant potential in terms of encompassing both linear and nonlinear dynamics in financial time series (Rundo et al., 2019). Nevertheless, the current hybrid studies do not have standardized protocols of integration, cohesive validation procedures, and market benchmarking (Sezer et al., 2020). Researchers are urged to come up with systematic hybrid modeling systems that are supported by reproducible assessment criteria and regimeconscious validation systems (Makridakis et al., 2018). like transformer- Moreover, new architectures, based time-series models, generative learning systems, have a potential to capture long-range dependencies and presence of multi-source information flows (Goodfellow et al., 2016). The predictive uses of alternative datasets, such as macroeconomic variables, news emotion, and social media the indicators, could predictive capacity and situational awareness (Gu et al., 2020). However, this integration should be supported by strict out-of-sample validation, rolling-window testing, and diagnostics of their robustness to reduce overfitting and guarantee generalizability (Lopez de Prado, 2018). The other research need is an interpretability. The use of explainable artificial intelligence (XAI) methods is needed to enhance transparency and theoretical insight on the role of features in deep learning architectures as they become more and more complex (Sezer et al., 2020). The integration of interpretability forecasting models will enable scholarly understanding and acceptance of the regulations. frameworks increase in the 4.2 Implications for Practice To financial practitioners, the review provides actionable insights of interest in the development of the trading strategies, optimization of the portfolio, in risk management, and in decision support systems (Gu et al., 2020). It has been shown that LSTM and other deep learning systems are especially efficient in sequential data modeling, which is why they can be used to predict intraday and volatility and in high-frequency trading settings (Fischer and Krauss, 2018). Nevertheless, the adoption of ML should be pursuit of caution on the part of the practitioners. Excel statistics is not always associated with profitable trading (Makridakis et al., 2018). In order to implement it practically, it is necessary to explicitly think about transaction costs, liquidity the compliance with issues, slippage, and regulations (Lopez de Prado, 2018). Thus, ML models must be integrated into cost-sensitive and risk-adjusted assessment systems instead of being assessed by measuring errors, e.g. RMSE or MAE (Sezer et al., 2020). Also, is dependent on interpretability. Banks are in need of more open that systems can provide decision-support explanations and stakeholders, to regulators of the predictive outputs (Gu et al., 2020).

  A SYSTEMATIC LITERATURE REVIEW AND BIBLIOMETRIC ANALYSIS OF MACHINE LEARNING ALGORITHMS AND TECHNICAL INDICATORS TO STOCK PRICE PREDICTION · 2026 · DOI

• In conclusion, RL offers a robust framework for optimizing portfolios by adapting to constantly changing market conditions while embedding risk management in deci- sion-making processes. Use of RL by investors leads to better risk-adjusted returns, making it an essential component of modern portfolio management. Although the proposed RL-based portfolio optimization model shows strong performance, it does have its drawbacks. The dependence of the model on historical data may hinder its capacity to respond to unexpected market situations, such as geopolitical conflicts or health crises. Moreover, the high computational demands for training and peri- odic updates can be excessive, particularly for extensive portfolios. Additionally, the existing model does not take into account further factors, such as news sentiment or macroeconomic indicators, which could offer important insights for decision- making. Future research could tackle these shortcomings by incorporating advanced data resources, utilizing online learning methods, and improving risk management through multi-objective optimization. Furthermore, enhancing the model’s explain- ability and scalability would make it more feasible for real-world applications. These improvements would not only boost the performance of the model but would also broaden its applicability across different financial markets and asset classes. Funding No funding received.

  Proximal Policy Optimization for Risk-Adjusted Stock Portfolio Optimization: A Reinforcement Learning Approach · 2026 · DOI

• The study identifies several shortcomings that should be addressed when interpreting the results beyond the growing role of artificial intelligence technologies in the financial markets and investment management practices, including 1) second-hand data is utilized from the academic journals, financial industry reports and publicly accessible financial databases instead of firsthand data collected from investment firms, portfolio managers, or financial institutions directly involved in the implementation of artificial intelligence technologies, and therefore the study findings are based on a significant amount of literature review, conceptual discussions and documented case studies but limited on real time operational evidence potentially limiting the breadth of the study in terms of reflecting on the full complexity of practical implementation of artificial intelligence technologies in portfolio management environments (Goodell et al., 2021; Bahoo et al., 2024); 2) The relatively limited empirical testing of artificial intelligence models in the research also imposes limitations as the study focuses largely on reviewing and synthesizing existing literature on AI applications in finance instead of conducting considerable experimental or quantitative testing of machine learning algorithms, predictive models, or automated trading systems using real financial data samples, preventing an assessment of the actual effectiveness, accuracy, and robustness of AI-driven investment strategies in different market conditions and economic scenarios (Giglio et al., 2022; Jiang et al., 2023); 3) The rapid technological environment in which the present study is utilizing methods of different analytic techniques, computational capabilities and investment management tools that may change quickly and make certain research findings or technological applications pointless as new and more advanced AI models and financial technologies will be introduced into the market, posing challenges for researchers who are trying to provide a long-term assessments of the effectiveness of AI systems (Deloitte, 2023; McKinsey Global Institute, 2023), and 4) Finally, there are some broader external influences on the rapidly integrated developments of artificial intelligence into financial markets which includes regulatory formulations, data accessibility, technological infrastructure, and organizational readiness in financial institutions, and these factors affecting the application of AI technologies were not explored extensively in the present study, and therefore represent a possible area for future research efforts in order to provide a more in-depth understanding on the practical aspects and real world effectivity of AI driven investment management systems in different financial market environments; thus, the study contributes to the expanding body of literature on the use of artificial intelligence in the financial sector, whilst further research opportunities are needed for future empirical validations, longitudinal studies and practical experimental studies to support and broaden the findings of the current study.

  The Role of Artificial Intelligence in Portfolio Management and Investment Decision-Making · 2026 · DOI

• Technical indicators selected through correlation analysis and domain knowledge may not include all relevant elements, thereby limiting model efficacy [4, 9]. While the models focus on maximizing cumulative profit, more advanced risk management measures are not employed, which may be required for real-world applications. RL models in a real-time trading environment face latency and execution speed problems. Ensuring that the models make timely judgments is critical for practical deployment. Further research will improve the trading agent by combining sentiment analysis from news and social media with macroeconomic variables to produce a more accurate understanding of market performance. In this regard, it is an intriguing avenue for future research to incorporate more dynamic reward functions and additional risk metrics such as VaR or CVaR in risk measures into the agent's decision-making process regarding profitability and risk management. Hybrid models that combine RL with additional methodologies such as time series analysis and hierarchical RL will aid in improving trading performance when applied to difficult scenarios. Low latency and scalability model optimization will enable effective real-time trading. Finally, expanding the technique to portfolio management and subjecting it to stress tests would increase the agent's robustness, as will its inherent risk management.

  Empirical analysis of Q learning, DQN and DDQN for intraday trading · 2026 · DOI

• Future work will focus on validating the proposed framework on larger, multi-plat- form, and cross-country datasets to assess scalability and external validity, while also incorporating drift-detection and robustness mechanisms to address sentiment instabil- ity, platform-specific bias, and evolving user behavior, incorporating rolling-window val- idation and walk-forward analysis on extended datasets.

  Application of deep learning-based social media data analysis in macroeconomic forecasting · 2026 · DOI

• Advantages of proposed SSLstacked LSTM Captures multi-temporal dependencies through stacked LSTM layers; models macroeconomic variables and sentiment indices jointly for improved forecasting accuracy…

  Application of deep learning-based social media data analysis in macroeconomic forecasting · 2026 · DOI

• Despite its contributions, the study has certain limitations. First, the findings are based on self-reported perceptions of investors and finance professionals, which may be subject to response bias and social desirability effects. Second, the cross-sectional nature of the study restricts the ability to capture changes in perceptions and adoption behaviour over time as AI technologies continue to evolve rapidly. Third, the study focusses primarily on perceived opportunities and constraints rather than objective performance outcomes, which may limit the generalizability of the results to actual investment performance. Fourth, contextual factors such as regulatory environment, technological infrastructure, and market maturity may vary across regions, thereby limiting the applicability of the findings beyond the study setting. Finally, the study does not differentiate extensively among types of AI tools or investment instruments, which may influence perceptions and adoption patterns differently.

  Integrating Artificial Intelligence into Financial Investment Decision-making: Opportunities and Constraints · 2026 · DOI

• The raw option strategy shows statistical significance only at the 5% confidence level (t-statistic of 2.25) with high volatility (annualized standard deviation 569.29%) and low Sharpe ratio (0.131 annualized). The sources of this extreme volatility and the conditions under which the LSTM IV forecasts fail to outperform random predictions remain uncharacterized.

  Predicting option prices from their price history via machine learning · 2026 · DOI

• The paper removes options with ex-dividend dates within the last week of expiry to mitigate early exercise effects in American options, but reports the empirical implications as 'very small' without detailed quantification. The specific magnitude of return distortion from early exercise risk in machine learning IV predictions requires explicit measurement.

  Predicting option prices from their price history via machine learning · 2026 · DOI

• The trading signal is derived from standardized ATM IV forecasts rather than the cross-sectional distribution of IVs at a given date, departing from Goyal and Saretto (2009). The relative performance of time-series versus cross-sectional IV prediction signals from the convolutional LSTM model has not been directly compared.

  Predicting option prices from their price history via machine learning · 2026 · DOI

• Portfolio construction uses value-weighting by dollar open interest to reflect market depth, but the robustness of this weighting scheme for thinly-traded option contracts during volatile market periods (e.g., 2018) is not analyzed. The comparison between open-interest weighting and alternative schemes for extreme market conditions is missing.

  Predicting option prices from their price history via machine learning · 2026 · DOI

• The delta-hedged trading strategy does not rebalance the delta hedge during the weekly holding period following the conservative approach of Goyal and Saretto (2009). The performance improvement from daily or intra-week delta rebalancing for machine learning-predicted IV forecasts has not been evaluated.

  Predicting option prices from their price history via machine learning · 2026 · DOI

• The paper uses a joint model specification across all underlying stocks rather than optimizing separate LSTM architectures for each individual stock. The impact of stock-specific model customization on IV forecasting accuracy and subsequent option trading strategy profitability remains unexplored.

  Predicting option prices from their price history via machine learning · 2026 · DOI

• This review has argued that AI in �inancial decision-making should be understood through a broader architecture rather than through isolated technical applications. The �irst core line is market prediction, where models attempt to extract tradable signals from prices, returns, volatility, and text. The second core line is �irm-level feature identi�ication, where models infer �inancing constraints, ESG characteristics, distress risk, and other latent corporate attributes. These lines have evolved separately, but they increasingly interact in real �inancial decisions. The literature on forecasting shows that the movement from classical econometric models to LSTM, transformers, and hybrid systems has expanded representational power, but not eliminated the basic problems of weak signals and unstable regimes (Sezer et al., 2020; Fischer & Krauss, 2018; Zhou et al., 2021). The portfolio literature shows that predictive gains matter only when they survive the translation into robust allocation under noisy covariance estimates, transaction costs, and rebalancing constraints (Markowitz, 1952; Kolm et al., 2014; Ban et al., 2018). The sustainable �inance literature shows that ESG and �inancing information now affect access to capital, cost of capital, risk, and resilience, but also suffer from substantial measurement disagreement and institutional heterogeneity (Cheng et al., 2014; Berg et al., 2022; Christensen et al., 2022). The two focal papers included by the user are therefore best interpreted not as isolated cases, but as representative nodes in this larger architecture. The LSTM-based portfolio paper captures the prediction-to-allocation logic that de�ines much of quantitative asset management (Li & Liu, 2023). The �inancing-constraint-and-ESG paper captures the �irm-side logic in which �inancial frictions and sustainability outcomes are jointly analyzed (Liu, 2022). The real research opportunity lies in connecting these domains.

  Artificial Intelligence in Financial Decision-Making · 2026 · DOI

Most-cited papers in Stock Market Forecasting Methods

• Overreaction in Macroeconomic Expectations · American Economic Review · 2020 · 537 citations
• Autoencoder asset pricing models · Journal of Econometrics · 2020 · 394 citations
• Deep Learning in Asset Pricing · Management Science · 2023 · 340 citations
• Bayesian BILSTM approach for tourism demand forecasting · Annals of Tourism Research · 2020 · 218 citations
• Wholesale price forecasts of green grams using the neural network · Asian Journal of Economics and Banking · 2024 · 205 citations
• A novel text-based framework for forecasting agricultural futures using massive online news headlines · International Journal of Forecasting · 2020 · 142 citations
• Artificial Intelligence and Strategic Decision-Making: Evidence from Entrepreneurs and Investors · Strategy Science · 2024 · 124 citations
• The role of text-extracted investor sentiment in Chinese stock price prediction with the enhancement of deep learning · International Journal of Forecasting · 2020 · 109 citations
• Nowcasting GDP using machine-learning algorithms: A real-time assessment · International Journal of Forecasting · 2020 · 101 citations
• Forecasting emerging technologies using data augmentation and deep learning · Scientometrics · 2020 · 97 citations

Most recent work

• Artificial Intelligence–Powered (Finance) Scholarship · Journal of Economic Literature · 2026
• HQNN-FSP: A hybrid classical-quantum neural network for regression-based financial stock market prediction · Quantum Machine Intelligence · 2026
• AVALIAÇÃO DE MODELOS HÍBRIDOS DE ANÁLISE TÉCNICA E FUNDAMENTALISTA: UMA PERSPECTIVA TECNOLÓGICA NO MERCADO DE AÇÕES · Zenodo (CERN European Organization for Nuclear Research) · 2026
• Análise e Visualização de Estratégias de Swing Trade com Streamlit e Plotly · Zenodo (CERN European Organization for Nuclear Research) · 2026
• DESENVOLVIMENTO DE UM SISTEMA MODULAR PARA PROSPECÇÃO E ANÁLISE DE DADOS DA B3 · Zenodo (CERN European Organization for Nuclear Research) · 2026
• Political Uncertainty and Credit Risk: The Role of Event Markets in Forecasting Ukraine's Sovereign Spreads · Scottish Journal of Political Economy · 2026
• Stock market price prediction using intelligence herpestidae algorithm with modified triplet attention-based hybrid deep learning model · SN Business & Economics · 2026
• Scalper Major: A Computational Solution for Automated Trading in High Volatility Environments · 2026
• Boost High Frequency Trading with Deep Reinforcement Learning and Transformer · Applied and Computational Engineering · 2026
• Artificial Intelligence for Modeling Investor Behavior and Market Sentiment Dynamics · International Journal of Artificial Intelligence Research · 2026

Related topics in Decision Sciences