What is the Fowlkes-Mallows Index?
The Fowlkes-Mallows Index (FMI, Fowlkes-Mallows Score or G-Mean) is a performance metric to evaluate the similarity of clusters obtained through various clustering algorithms. It is typically used to evaluate the clustering performance of a specific algorithm by assuming that the cluster it is compared to is the ground truth — i.e. the perfect cluster.
With this in mind, one can easily grasp how it is also a valid (external) measure for evaluating supervised learning algorithms. The FMI is defined as the geometric mean of the pairwise precision and recall.
The FMI has several advantages:
- With a random classification, the FMI will approach zero.
- The FMI is upper-bound. A perfect classification will result in an FMI of 1.
Further reading: