CLASSify gives researchers a way to train tabular data automatically on many machine-learning models and find which produces the best results, without programming.
The interface reports explainability scores for each feature, showing how each column of the data affects the model’s predictions. CLASSify also generates synthetic data that follows the patterns of the real data, for datasets with imbalanced classes or protected information.
Part of the Kentucky Open Science platform: it runs over the Cresco mesh, with versioned storage for datasets, models, and code in GaiaKeep. See the full stack.
Ten supervised models: random forest, gradient boosting, histogram-based gradient boosting, XGBoost, bagging, logistic regression, SGD classifier, k-nearest neighbors, multi-layer perceptron, and TabPFN.
Three unsupervised methods: k-means, spectral clustering, and HDBSCAN.
Each model accepts custom parameters, defaults, or automatic tuning with Optuna. Explainability scores use SHAP, and synthetic data uses the Synthetic Data Vault library.
CLASSify is available on an individual basis after an administrator grants access. Request access through the information request form. A tutorial video is available.
A. D. Mullen, S. E. Armstrong, J. Talbert, V. K. C. Bumgardner. CLASSify: A Web-Based Tool for Machine Learning. AMIA Joint Summits on Translational Science Proceedings, 2024, pp. 364-373. PubMed