forester: A Novel Approach to Accessible and Interpretable AutoML for Tree-Based Modeling

Anna Kozak, Hubert Ruczyński

Keywords: machine learning, automated machine learning, tree-based models, automated reporting

TL;DR: The AutoML package in R is designed for tabular data regression and binary classification tasks, which focuses on the simplicity of use and interpretability of the results.

Abstract:

The majority of AutoML solutions are developed in Python. However, a large percentage of data scientists are associated with the R language. Unfortunately, there are limited R solutions available with high entry level which means they are not accessible to everyone. To fill this gap, we present the $\textit{forester}$ package, which offers ease of use regardless of the user's proficiency in the area of machine learning.

The $\textit{forester}$ package is an open-source AutoML package implemented in R designed for training high-quality tree-based models on tabular data. It supports regression and binary classification tasks. A single line of code allows the use of unprocessed datasets, informs about potential issues concerning them, and handles feature engineering automatically. Moreover, hyperparameter tuning is performed by Bayesian optimization, which provides high-quality outcomes. The results are later served as a ranked list of models. Finally, the $\textit{forester}$ package offers a vast training report, including the ranked list, a comparison of trained models, and explanations for the best one.

Abcd Fit: Applications

Submission Checklist: Yes

Broader Impact Statement: Yes

Paper Availability And License: Yes

Code Of Conduct: Yes

CPU Hours: 0

GPU Hours: 0

TPU Hours: 0