Improving ML Model Accuracy Through Data-Centric Engineering in Enterprise Data Lakes

This research explores the significance of data-centric engineering in improving the accuracy of machine learning (ML) models, particularly within the context of enterprise data lakes. While traditional approaches to enhancing ML model performance have predominantly focused on refining algorithms and model architectures, this study shifts the focus to data quality as a key driver of model success. The paper reviews various studies that demonstrate how data-centric practices—such as data cleaning, augmentation, labeling, bias correction, and data provenance tracking—can substantially improve the performance and reliability of ML models. The research also highlights the challenges faced by organizations in managing large-scale, diverse datasets within enterprise data lakes, and how these challenges can be overcome through systematic data management strategies. The findings emphasize that prioritizing data quality over model complexity not only boosts model accuracy but also enhances fairness, interpretability, and scalability, ultimately leading to more efficient and ethical ML systems.