Altair RISE
A program to recognize and reward our most engaged community members
Nominate Yourself Now!
Home
Discussions
Community Q&A
labeled DBSCAN clustering with map on labels, how to easily find working epsilon/min points?
pvds90
Hi guys,
I have 1 labeled attribute and i want to use DBSCAN. Is there a way how i can easily find what epsilon and min points are a fit? Now I'm trying randomly, but this is very time consuming.
Find more posts tagged with
AI Studio
Clustering
Accepted answers
jacobcybulski
The easiest way is to place the clustering operator together with a number of "daisy-chained" cluster performance operators within an Optimize Parameters (Grid). The optimizer will allow you to vary DBSCAN parameters, such as epsilon and the minimum points, while performance is collected and logged for each combination of these parameters. At the end (and while it is executing) you can watch and plot the logged values. Note that DBSCAN is slow on very large data sets, so I suggest to take a smaller data sample and initially vary the parameters in large steps. Once you find the best "cube" of your parameters, tune the parameters with the finer comb.
All comments
jacobcybulski
The easiest way is to place the clustering operator together with a number of "daisy-chained" cluster performance operators within an Optimize Parameters (Grid). The optimizer will allow you to vary DBSCAN parameters, such as epsilon and the minimum points, while performance is collected and logged for each combination of these parameters. At the end (and while it is executing) you can watch and plot the logged values. Note that DBSCAN is slow on very large data sets, so I suggest to take a smaller data sample and initially vary the parameters in large steps. Once you find the best "cube" of your parameters, tune the parameters with the finer comb.
Quick Links
All Categories
Recent Discussions
Activity
Unanswered
日本語 (Japanese)
한국어(Korean)
Groups