Regularized target encoding outperforms traditional methods in supervised machine learning with high cardinality features

• Published in: Computational statistics Vol. 37; no. 5; pp. 2671 - 2692
• Main Authors: Pargent, Florian, Pfisterer, Florian, Thomas, Janek, Bischl, Bernd
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.11.2022; Springer Nature B.V
• Subjects: Algorithms; Best practice; Data analysis; Economic Theory/Quantitative Economics/Mathematical Methods; Machine learning; Mathematics and Statistics; Original Paper; Performance prediction; Probability and Statistics in Computer Science; Probability Theory and Stochastic Processes; Statistics; Support vector machines; Benchmark; Dummy encoding; Supervised machine learning; Generalized linear mixed models; Target encoding; High-cardinality categorical features
• ISSN: 0943-4062; 1613-9658
• DOI: 10.1007/s00180-022-01207-6

Since most machine learning (ML) algorithms are designed for numerical inputs, efficiently encoding categorical variables is a crucial aspect in data analysis. A common problem are high cardinality features, i.e. unordered categorical predictor variables with a high number of levels. We study techniques that yield numeric representations of categorical variables which can then be used in subsequent ML applications. We focus on the impact of these techniques on a subsequent algorithm’s predictive performance, and—if possible—derive best practices on when to use which technique. We conducted a large-scale benchmark experiment, where we compared different encoding strategies together with five ML algorithms (lasso, random forest, gradient boosting, k -nearest neighbors, support vector machine) using datasets from regression, binary- and multiclass–classification settings. In our study, regularized versions of target encoding (i.e. using target predictions based on the feature levels in the training set as a new numerical feature) consistently provided the best results. Traditionally widely used encodings that make unreasonable assumptions to map levels to integers (e.g. integer encoding) or to reduce the number of levels (possibly based on target information, e.g. leaf encoding) before creating binary indicator variables (one-hot or dummy encoding) were not as effective in comparison.