building-baseline-models

What problem does it solve?

This Skill prevents you from judging a machine-learning model in isolation by forcing a fair baseline comparison first, so you can tell whether a complex model is truly better than chance or a simpler alternative.

Core Features & Use Cases

• Baseline ladder: Recommends dummy, linear, and tree-based baselines before complex models such as XGBoost, LightGBM, neural nets, or transformers.
• Fair evaluation: Requires the same train/test split or cross-validation folds, the same metric, and the same preprocessing pipeline across every model.
• Decision support: Flags cases where a reported score is not meaningful without baseline context and refuses to certify a complex model as good when the evidence is insufficient.
• Use case: If you are about to report an XGBoost ROC-AUC, this Skill tells you which baselines to fit first and how to compare them responsibly.

Quick Start

Ask for a baseline ladder for your supervised learning task and have the skill compare dummy, linear, and random forest models on the same data split and metric before you report your final model.