Data-Driven Intelligent Systems: A Survey of Representation Learning, Predictive Modeling, Causal Reasoning, and AI-Driven Decision Optimization
- Authors
-
-
Liora Veyne
Department of Computer Science, Lakehead University, Thunder Bay, Canada
Author
-
- Abstract
-
Data-driven intelligent systems have become an important research direction in modern artificial intelligence because they connect large-scale data processing, representation learning, predictive modeling, causal reasoning, and adaptive decision optimization into a unified computational workflow. Different from traditional data analytics, which often emphasizes descriptive statistics or isolated prediction, recent data-driven systems are designed to learn from heterogeneous, dynamic, noisy, and high-dimensional data while supporting decisions under uncertainty. This survey reviews recent studies on data-driven intelligent systems and synthesizes the literature from a methodological perspective rather than from a domain-by-domain perspective. The reviewed works are organized around representation learning, sequential and spatiotemporal prediction, counterfactual and causal reasoning, anomaly and risk modeling, large language model agents, and system-level challenges. The survey shows that Transformer models, graph neural networks, contrastive learning, continual learning, retrieval-augmented modeling, and agentic large language model frameworks are becoming key foundations for robust data-driven intelligence. Meanwhile, several limitations remain unresolved, including data heterogeneity, distribution shift, interpretability, causal validity, evaluation beyond accuracy, and governance of automated decision pipelines. The paper concludes by identifying future directions toward unified representation-reasoning frameworks, drift-aware decision systems, auditable AI pipelines, human-in-the-loop decision coordination, and more comprehensive benchmarks for evaluating data-driven intelligent systems.
- Downloads
- Published
- 2026-04-01
- Issue
- Vol. 6 No. 4 (2026)
- Section
- Articles