Explain unsupervised machine learning algorithms (k-means clustering, principal components analysis and hierarchical clustering) and decide which kinds of problem each handles best
The idea
Principal components analysis replaces many correlated features with a smaller set of uncorrelated composites. Each component's eigenvalue measures the variance it explains; divide by the sum of eigenvalues for its share, and keep enough components to reach a chosen threshold, often 85% to 95%, using a scree plot. The trade-off is that components are hard to interpret. k-means clustering assigns observations to k clusters, with k fixed in advance, by repeatedly moving centroids. Hierarchical clustering builds a tree of clusters, bottom up (agglomerative) or top down (divisive), shown as a dendrogram, so the number of clusters can be chosen afterwards. PCA reduces features; clustering groups observations.
In the app: 2 practice questions on this outcome, a narrated walkthrough with diagrams, spaced review that brings it back before you forget it.
Take the free diagnostic →More in Machine Learning
- Explain what deep learning is, and what machine learning is in its supervised and unsupervised forms
- Explain overfitting and recognise ways to deal with it
- Explain supervised machine learning algorithms (support vector machine, penalized regression, k-nearest neighbor, random forest, classification and regression tree, and ensemble learning) and decide which kinds of problem each handles best