RuleXAI—A package for rule-based explanations of machine learning model

Abstract

The ability to use eXplainable Artificial Intelligence (XAI) methods is very important for both AI users and AI developers. This paper presents the RuleXAI library, which provides XAI methods based on rule-based models. The package presented can be applied to classification, regression and survival analysis tasks. RuleXAI operates on elementary rule conditions and enables the generation of global explanations, local explanations and the generation of a new data representation, simplifying data preprocessing. The explanations of model decisions that are generated by RuleXAI rely on feature relevance and provide information not only about the importance of attributes, but also about the importance of attribute values.

Keywords

• XAI
• Rule-based representation
• Global explanations
• Local explanations
• Feature relevance
• Python