To impute or not to impute: How machine learning modelers treat missing data
This addresses a critical gap in ML practice by highlighting poor decision-making in missing data treatment, which is incremental as it identifies an existing problem without proposing new solutions.
The study surveyed 70 ML researchers and engineers to understand how they handle missing data in tabular models, finding that most make uninformed decisions that could compromise model validity.
Missing data is prevalent in tabular machine learning (ML) models, and different missing data treatment methods can significantly affect ML model training results. However, little is known about how ML researchers and engineers choose missing data treatment methods and what factors affect their choices. To this end, we conducted a survey of 70 ML researchers and engineers. Our results revealed that most participants were not making informed decisions regarding missing data treatment, which could significantly affect the validity of the ML models trained by these researchers. We advocate for better education on missing data, more standardized missing data reporting, and better missing data analysis tools.