ShinyLearner

ShinyLearner performs systematic benchmarking of machine-learning classification algorithms to enable empirical comparison and selection of classifiers for life-science datasets.

Key Features:

• Integration of Multiple Machine-Learning Packages: Integrates multiple machine-learning packages into software containers to standardize execution across libraries and implementations.
• Multi-algorithm, Multi-dataset Benchmarking: Applies multiple classification algorithms to various datasets to evaluate overall performance trends and algorithm comparisons.
• Nested Cross-Validation for Hyperparameter Optimization and Feature Selection: Supports hyperparameter optimization and feature selection via nested cross-validation while tracking operations within nested processes.
• Transparent Output Generation: Generates comprehensive output files documenting hyperparameter tuning, feature selection steps, and benchmarking results to support reproducibility.

Scientific Applications:

• Algorithm selection for life-science datasets: Empirically compares classifiers to identify models that improve classification accuracy on biological and life-science data.
• Comparative evaluation across software packages: Enables comparisons of algorithms implemented in different libraries and implementations to assess implementation-dependent differences.
• Model and feature optimization: Supports selection of optimal hyperparameters and feature subsets for complex biological datasets.

Methodology:

Integrates machine-learning packages into software containers, applies multiple classification algorithms to various datasets, performs hyperparameter optimization and feature selection using nested cross-validation while tracking nested operations, and generates comprehensive output files documenting each benchmarking step and hyperparameter combination.