This example simulates a multi-label document classification problem. The dataset is generated randomly based on the following process:
In the above process, rejection sampling is used to make sure that n is more than 2, and that the document length is never zero. Likewise, we reject classes which have already been chosen. The documents that are assigned to both classes are plotted surrounded by two colored circles.
The classification is performed by projecting to the first two principal components found by PCA and CCA for visualisation purposes, followed by using the sklearn.multiclass.OneVsRestClassifier metaclassifier using two SVCs with linear kernels to learn a discriminative model for each class. Note that PCA is used to perform an unsupervised dimensionality reduction, while CCA is used to perform a supervised one.
Note: in the plot, “unlabeled samples” does not mean that we don’t know the labels (as in semi-supervised learning) but that the samples simply do not have a label.