Data-Centric Artificial Intelligence: Improving Model Performance through Quality-Driven Data Engineering

Abstract

This study examined the transformative potential of Data-Centric Artificial Intelligence (DCAI) in improving model performance through quality-driven data engineering. Moving beyond algorithmic refinement, the research focused on how enhanced data quality—achieved through systematic cleaning, annotation, augmentation, and validation—contributed to the robustness, fairness, and interpretability of AI models. A mixed-methods approach was employed, combining experimental evaluation of machine learning models across multiple datasets with qualitative insights into data governance and lifecycle management. The findings revealed that high-quality data substantially increased predictive accuracy, reduced overfitting, and improved model generalization across domains. Moreover, the integration of automated data validation tools and human-in-the-loop systems enhanced dataset reliability and minimized bias. The study also emphasized the importance of ethical data sourcing and transparency, aligning with emerging global AI governance frameworks. It concluded that prioritizing data quality over algorithmic complexity produced more sustainable and trustworthy AI systems. The research provided actionable recommendations for embedding data governance, automation, and interdisciplinary collaboration within AI pipelines. Future directions included developing standardized data quality metrics, exploring explainable AI integration, and leveraging federated learning and synthetic data to scale data-centric frameworks. This paradigm shift positioned data as the foundational element driving the next generation of reliable and ethical artificial intelligence.

Ähnliche Arbeiten