Elbow method, interpreting graph, how many clusters

Dear Knime Fellows
I’m learning KNIME, feeling very insecure about my interpretation.
For LDA topic detection, I’m trying to determine the correct/ideal number of clusters (k).
I got the scatter plot, to me the k (or elbow) point looks to be 7 - am I correct?
but what confuses me is that value for k=8 is actually higher: is this an indicator of a mistake I have made?
many thanks for your help!
Outi

Hello outiloni,

The best option of K is where the SSE decreases the most. In this case, I believe k=7 is the optimal solution since k=6 has the first major drop.

You can read more here: https://www.knime.com/blog/topic-extraction-optimizing-the-number-of-topics-with-the-elbow-method

4 Likes

Thank you so much Jinweisun!!! Very much appreciated!!!
Outi

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.