A python library for data-driven causal fuzzy classification rule generation -- mablars

• Published in: IEEE International Fuzzy Systems conference proceedings pp. 1 - 6
• Main Authors: Zhang, Te, Wagner, Christian
• Format: Conference Proceeding
• Language: English
• Published: IEEE 06.07.2025
• Subjects: Causal Rules; Classification algorithms; Computational modeling; Fuzzy; Fuzzy sets; Fuzzy systems; Libraries; Natural languages; Open source software; Python; Software tools; Source coding; XAI
• ISSN: 1558-4739
• DOI: 10.1109/FUZZ62266.2025.11152225

A principal focus in fuzzy systems research is on maintaining good performance while providing strong explainability, principally by leveraging meaningful sets of human-accessible rules. In practice, while a variety of software tools have been developed to facilitate implementing fuzzy systems, none of them focus on generating rules for fuzzy systems which specifically capture causal relationships between the system variables. However, rules which reflect causal relationships between variables are critical as a meaningful step towards rule and system explainability. In previous works, a causal rule generation framework for fuzzy systems called Markov blanket rule generation framework (MABLAR) was established. In this paper, we introduce mablars which is a Python library implementing the MABLAR framework and making it freely and easily accessible. mablars supports the implementation of different MABLAR variants for classification problems. We provide an overview of the current feature set of the open-source software package, and demonstrate it using a publicly available data set. Finally, we include an overview of the computational complexity-in practice-of the different variants of MABLAR supported thus far, to further aid adopters to choose a model appropriate to their needs.