KAXAI: An Integrated Environment for Knowledge Analysis and Explainable AI

Democratizing Machine Learning Through User-Friendly Design: KAXAI demonstrates that complex ML systems can be made accessible to non-technical users by integrating AutoML (automated machine learning), explainable AI, and synthetic data generation into a single platform that abstracts technical complexity while maintaining high usability.

Novel Interpretability Method Outperforms Existing Approaches: The paper introduces MEDLEY, a new model-dependent local interpreter that was evaluated against established methods (LIME, Greedy, and Parzen), suggesting improvements in how AI systems can explain their decisions to users—a critical component of responsible AI.

GAN-Enhanced Data is Most Reliable for Synthetic Data Generation: When comparing three synthetic data approaches (LLM-based, library-based, and GAN-enhanced), the research found that enhancing original datasets with GANs produced the most reliable synthetic data according to statistical tests (KS tests and standard deviation analysis), with particular effectiveness for quantitative datasets.

High Accuracy Achieved with Interpretable Models: The proposed classifiers (Logistic Regression Forest and Support Vector Tree) achieved strong performance—96% accuracy on a diabetes dataset and 93% on a survey dataset—demonstrating that interpretability and performance need not be sacrificed for one another.

Practical Path Forward for Responsible AI Implementation: By combining automated model selection, built-in explainability tools, and synthetic data capabilities in one integrated system, KAXAI provides a concrete framework for organizations to deploy machine learning responsibly without requiring deep technical expertise.