- Learn
- Machine Learning
- Classification
ML-M04
Classification
Predicting categories: logistic regression, nearest neighbours, Naive Bayes, SVMs, and the metrics and calibration that make their outputs trustworthy.
Classification is the other half of supervised learning: instead of a number, the model picks a category, such as spam or not spam, fraud or legitimate, cat, dog or bird. This module covers four classic families that still run a large share of production systems: logistic regression, k-nearest neighbours, Naive Bayes and support vector machines. Each one draws its decision boundary in a different way, and you will drag points and watch those boundaries move.
The second half is about judging and trusting a classifier: the confusion matrix and the metrics built from it, ROC and precision-recall curves, probability calibration, and how all of this extends from two classes to many classes and to several labels at once.
Lessons
- ML-04.01 Logistic regression Logistic regression predicts a probability for a binary outcome by passing a linear score through the sigmoid function, which squashes any number into the range 0 to 1. Beginner 15 min
- ML-04.02 k-Nearest Neighbours k-nearest neighbours classifies a new point by looking at the k closest training points and taking a majority vote. Beginner 10 min
- ML-04.03 Naive Bayes Naive Bayes applies Bayes' theorem with the simplifying assumption that features are independent given the class. Intermediate 10 min
- ML-04.04 Support Vector Machines and kernels A support vector machine finds the boundary with the widest margin between classes; only the points closest to the boundary, the support vectors, define it. Intermediate 15 min
- ML-04.05 Classification metrics A confusion matrix counts true and false positives and negatives. Beginner 16 min
- ML-04.06 Probability calibration A calibrated model's probabilities mean what they say: of all predictions at 0.8, about 80 percent are positive. Advanced 10 min
- ML-04.07 Multi-class and multi-label classification Multi-class problems pick one class out of many and usually use a softmax output that produces probabilities summing to 1. Intermediate 10 min