Tomaz Kastrun shows how to use R to find clusters of queries which behave similarly:

So the R code said that, there are three clusters generating And I used medians to generate data around it. In addition I have also tested the result with Partitioning around medoids (which is opposite to hierarchical clustering) and the results from both techniques yield clean clusters.

Clustering models can be powerful for discovering commonalities, and that might help you find a number of queries which all behave in some sub-optimal way without having to trawl through every procedure’s code.

Comments closed