Learning path
Next moduleModule 5: Instance-based and probabilistic models
Module 4
Classification
Predict classes and calibrated probabilities with logistic regression, then turn scores into decisions with the right metric and threshold.
- 1. Logistic regressionTurn a linear score into a probability with the sigmoid, train it with cross-entropy, and extend it to many classes with softmax.35 min
- 2. Classification metrics and thresholdsBuild a confusion matrix, compute precision, recall, specificity, and F1, read ROC and PR curves, check calibration, and pick a threshold from the real cost of errors.40 min
Classification models output a score or probability; a threshold turns that score into an action. This module covers the model (logistic regression, sigmoid, softmax, cross-entropy) and then the part that decides real-world impact: confusion matrices, precision and recall, ROC and PR curves, calibration, and class imbalance.
| # | Lesson | You will be able to |
|---|---|---|
| 1 | Logistic regression | compute logits, probabilities, and cross-entropy by hand |
| 2 | Classification metrics and thresholds | build a confusion matrix, compute precision, recall, and F1, and choose a threshold from error costs |
Interactive lab: Classification threshold