Data clustering

A system for measuring similarity between a first set of clusters generated by a first clustering procedure and a set of object labels, the system comprising:

• a first computer-readable medium configured to store computer-executable instructions;
• and
• one or more processors configured to cause the system to:
• retrieve a set of objects from one or more computer-readable media,
• apply the first clustering procedure to the set of objects to sort the objects into a first set of clusters, wherein the first set of clusters has a first number of clusters,
• retrieve the set of object labels from the one or more computer-readable media,
• generate a similarity index between the first set of clusters and the set of object labels,
• generate an expected value of the similarity index, wherein the expected value of the similarity index indicates an average value of the similarity index between randomly generated sets of clusters and the set of object labels, wherein the randomly generated sets of clusters have at most a same number of clusters as the first number of clusters, and wherein the randomness of the randomly generated sets of clusters assumes that a set of first cluster sizes is different than a set of the randomly generated cluster sizes;
• and
• adjust the calculated similarity index by a penalty factor that is based at least in part on the expected value of the similarity index.

One or more non-transitory computer-readable media storing instructions that, when executed by one or more computing devices, cause the one or more computing devices to perform operations comprising:

• retrieving a set of objects from one or more computer-readable media;
• generating a first set of clusters based on a first clustering procedure and the set of objects;
• calculating a similarity index between the first set of clusters and a set of object labels;
• calculating an expected value of the similarity index, wherein the expected value of the similarity index is a weighted average of possible values of the similarity index between a randomly generated set of clusters and the object labels, wherein the randomness of the randomly generated set of clusters assumes that a set of first cluster sizes is different than a set of the randomly generated cluster sizes;
• and
• modifying the calculated similarity index by a penalty factor that is based at least in part on the expected value of the similarity index.