DSAverse

Initializing Sorting Algorithms...

Sorting

Trees

Graphs

Preparing interactive visualizations...

Back to Machine Learning

Classification Metrics

Turn a model's scores into yes or no, count the four kinds of outcome, and watch precision, recall, F1 and the ROC and precision–recall curves move as the threshold slides. Then meet the 95%-accurate test that finds nobody.

Builds onLogistic Regression Conditional Probability Bayes’ Theorem

Logistic regression fitted in your browser · seeded test sets

Act 1 of 6 · Scoresstep 1 of 24

Speed

120 test points: 30 actually positive (orange) and 90 negative (blue). A logistic regression fitted to 300 other points gives each one a score — its estimated chance of being positive. It never saw these 120. Positives tend to score higher, but the two groups overlap.

\hat y = \begin{cases} 1 & s \ge t \\ 0 & s < t \end{cases}

Classification Metrics

Builds onLogistic Regression Conditional Probability Bayes’ Theorem

Logistic regression fitted in your browser · seeded test sets

Act 1 of 6 · Scoresstep 1 of 24

Speed

\hat y = \begin{cases} 1 & s \ge t \\ 0 & s < t \end{cases}

DSAverse

Classification Metrics

See also

Classification Metrics

See also