DANTE: Hybrid AI System for Context-Aware Interpretable Feature Engineering
Year: 2025
Type: conference-paper
Abstract: Machine learning has become essential in data management, enabling accuracy and efficiency in handling complex datasets. For structured data, however, feature engineering is required to enhance predictive performance but remains a time-consuming, domain-specific task. Moreover, as ML systems are increasingly adopted, ensuring interpretability, particularly for domain experts, has become important. To address these challenges, we proposed KRAFT, an automated feature engineering (AutoFE) approach that combines a neural generator for feature transformation with a knowledge-based discriminator to ensure interpretability through symbolic reasoning. We have also empirically shown that KRAFT outperforms SOTA approaches in both accuracy and interpretability. In this demonstration, we introduce DANTE, an interactive tool based on KRAFT to facilitate feature generation and model evaluation on structured data. Its no-code interface empowers users to generate meaningful, interpretable features for real-world applications. A demonstration video of DANTE is available at https://DANTE.com.
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4411374267