Learning Data Mining with Python - Second Edition

Extracting topic information from clusters

Now we set our sights on the clusters in an attempt to discover the topics in each.

We first extract the term list from our feature extraction step:

terms = pipeline.named_steps['feature_extraction'].get_feature_names()

We also set up another counter for counting the size of each of our classes:

c = Counter(labels)

Iterating over each cluster, we print the size of the cluster as before.

It is important to keep in mind the sizes of the clusters when evaluating the results—some of the clusters will only have one sample, and are therefore not indicative of a general trend.

Next (and still in the loop), we iterate over the most important terms for this cluster. To do this, we take the five largest ...