Evaluating clustering results
The current issue of Bioinformatics starts of with the very useful review of the evaluation of clustering results.
Clustering is easy: You always get a result that can be called a success by one measure or the other. The authors of the review consider mathematical properties of the solutions and show how to compare them amongst each other.
It is a handy reference for your next clustering problem - with all the high (and medium) throughput data pouring out of the labs into the bioinformatics offices, there is a steady interest in such techniques. Even if clustering often fails to explain much of the data.
Clustering is easy: You always get a result that can be called a success by one measure or the other. The authors of the review consider mathematical properties of the solutions and show how to compare them amongst each other.
It is a handy reference for your next clustering problem - with all the high (and medium) throughput data pouring out of the labs into the bioinformatics offices, there is a steady interest in such techniques. Even if clustering often fails to explain much of the data.
spitshine - 2005-07-15 17:39