KnowFE : A Hybrid AI System for Explainable Feature Engineering using Knowledge-Guided Reinforcement Learning

Feature engineering is a critical yet often manual step in building effective machine learning models. While automated machine learning (AutoML) has streamlined many aspects of model development, the generation of high-quality, interpretable features remains a key bottleneck, requiring case-by-case domain knowledge and significant effort. This challenge highlights the importance of automated feature engineering (AutoFE) as a critical component within AutoML pipeline. To address this, we recently proposed SMART, a novel AutoFE approach that combines knowledge graph reasoning and deep reinforcement learning to guide the generation of interpretable features. In this demonstration, we introduce KnowFE, a web-based AutoFE platform powered by SMART. KnowFE enables users to generate high-quality, human-understandable features without writing any code, striking a balance between explainability and predictive performance. With a user-friendly interface, it empowers data practitioners to efficiently enhance machine learning workflows across diverse domains. A video demonstration is available at https://www.KnowFE.com .