Cost-Aware Data Quality Assessment and Remediation for Machine Learning

Description

This research project investigates how common data-quality problems affect the performance of machine-learning models and how these problems can be addressed efficiently. The project examines different types of errors, including missing values, outliers, duplicates, inconsistent data, and invalid values. The student will contribute to an experimental framework that evaluates the impact of these errors on predictive performance, measures the effectiveness of automated cleaning techniques, and compares their computational costs. The ultimate goal is to support evidence-based decisions about which data-quality problems should be addressed first to improve machine-learning performance while minimizing computational resources.