Missing Data Infill with Automunge
This work addresses missing data handling for tabular data preprocessing in data science, offering incremental improvements through automated ML-based imputation.
The paper benchmarks missing data imputation methods in the Automunge library, finding that ML infill often outperforms other methods for numeric and categoric features, with performance gains enhanced by adding support columns signaling infill presence.
Missing data is a fundamental obstacle in the practice of data science. This paper surveys a few conventions for imputation as available in the Automunge open source python library platform for tabular data preprocessing, including "ML infill" in which auto ML models are trained for target features from partitioned extracts of a training set. A series of validation experiments were performed to benchmark imputation scenarios towards downstream model performance, in which it was found for the given benchmark sets that in many cases ML infill outperformed for both numeric and categoric target features, and was otherwise at minimum within noise distributions of the other imputation scenarios. Evidence also suggested supplementing ML infill with the addition of support columns with boolean integer markers signaling presence of infill was usually beneficial to downstream model performance. We consider these results sufficient to recommend defaulting to ML infill for tabular learning, and further recommend supplementing imputations with support columns signaling presence of infill, each as can be prepared with push-button operation in the Automunge library. Our contributions include an auto ML derived missing data imputation library for tabular learning in the python ecosystem, fully integrated into a preprocessing platform with an extensive library of feature transformations, with a novel production friendly implementation that bases imputation models on a designated train set for consistent basis towards additional data.