A Hybrid AI-Driven Decision Support System for Retail Investors: A Comprehensive Survey of Technical, Fundamental, and Sentiment-Based Models

Abstract

Retail investors frequently encounter significant challenges in interpreting diverse and complex market signals, which include dynamic technical trends, nuanced financial ratios, and volatile sentiment-driven information. The concurrent advancements in Artificial Intelligence (AI), sophisticated Ma- chine Learning (ML) techniques, and robust real-time financial APIs have paved the way for the development of highly automated Decision Support Systems (DSS). This comprehensive survey reviews key research published between 2024 and 2025 across three foundational domains of financial forecasting: (i) Fundamental-Analysis-based models, (ii) Technical-Indicator- driven ML frameworks, and (iii) Deep Learning-based Sentiment Forecasting Systems. A detailed comparative analysis is provided, highlighting the performance metrics, architectural nuances, and practical usability strengths and limitations of each approach. Finally, we propose a novel, integrated Hybrid DSS Architecture. This architecture is designed to fuse the predictive power of all three data streams, supporting high- fidelity real-time prediction, enhanced retail usability, and critical Explainable AI (XAI) capabilities to build trust and transparency.

Introduction

Financial markets are influenced by quantitative indicators, macroeconomic policies, and investor sentiment. Retail investors often lack access to sophisticated tools, leading to suboptimal decisions, whereas AI advancements—Deep Learning, NLP, and high-frequency data—enable automated Decision Support Systems (DSS) to bridge this gap. Traditional DSS approaches focus on three streams: Technical Analysis (price and volume trends), Fundamental Analysis (financial health and ratios), and Sentiment Analysis (market mood from unstructured text). Each stream has strengths but also limitations: technical models may fail during market regime shifts, fundamental models are slow to react, and sentiment models are computationally heavy and prone to noise.

To address these limitations, a Hybrid DSS architecture is proposed, integrating all three analytical streams through feature fusion and ensemble learning. The system produces actionable investment signals (e.g., Buy, Sell) while incorporating Explainable AI (XAI) to provide transparency in decision-making. Key challenges include real-time deployment, model adaptability to non-stationary markets, and handling noisy sentiment data from social media.

Conclusion

The 2024–2025 research landscape shows significant progress in developing sophisticated financial forecasting models, with notable advancements in technical-indicator- driven ML (P2), rule-based fundamental analysis (P1), and deep learning sentiment processing (P3). However, a signif- icant gap exists in deploying a truly unified, real-time, and trustworthy decision support system for retail investors. The proposed Hybrid DSS Architecture addresses this by creating a robust Feature Fusion Layer that optimally weights the predictions from all three domains. The mandatory inclusion of an XAI Module ensures transparency and interpretability, transforming the AI from a black-box oracle into a trusted, explainable analytical partner, which is essential for mass retail adoption.

References

[1] Abrishami, A., et al. (2024). A Rule-Based Decision Support System using Fundamental Indicators for Portfolio Selection. arXiv preprint arXiv:2412.05297. [2] Mostafavi, S., and Hooman, A. (2025). High-Performance Stock Prediction using a Systematic Comparison of Technical Indi- cator Feature Sets and Ensemble Machine Learning. Elsevier Journal of Financial Data Science. [3] Tafara, R.H., et al. (2025). Contextualizing Financial Sentiment: Topic Modeling and Sequence Prediction with GRU/LSTM Architectures. Elsevier Journal of Decision Support Systems.

Copyright

Copyright © 2025 Komal Tambe, Swaroop Biradar, Nitesh Kumar Kushwaha. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.