evaluating-multiclass-classifiers

What problem does it solve?

This Skill helps you evaluate a trained multi-class classifier correctly instead of relying on misleading single-number summaries like overall accuracy.

Core Features & Use Cases

• Produces a complete evaluation report for single-label classifiers with three or more classes.
• Reports per-class precision, recall, F1, support, macro-F1, weighted-F1, micro-F1, confusion matrices, top-k accuracy, one-vs-rest ROC and PR metrics, and calibration checks.
• Flags class imbalance, highlights dominant confusions, and refuses to treat a binary task as multi-class.
• Use it for image classification, intent classification, ICD-code prediction, severity triage, and other long-tail classification problems.

Quick Start

Ask the Skill to evaluate your multiclass model using y_true and y_pred_proba and produce a full report with per-class metrics, confusion matrices, aggregation choice, and diagnostic warnings.