Learning from biological data
Supervised learning uses known targets to learn a prediction from features.
Loading video…
Supervised learning uses known targets to learn a prediction from features. Regression predicts a continuous measurement; classification predicts a category. Unsupervised methods explore structure without target labels. Define the biological unit of each sample before modeling, because repeated measurements from one patient are related rather than independent examples.
Worked example
To predict a continuous biomarker, rows are patients, features are baseline measurements and the target is a later biomarker value. A disease/control label would instead make this classification.