Error Sum of Squares (SSE) is the sum of the squared differences between each observation and its group's mean. It can be used as a measure of variation within a cluster. If all cases within a cluster are identical the SSE would then be equal to 0. Anyone can give an example of calculating SEE? since it seems my results will be far away from 0, while I see the results on the previous work are quite close to 0. How this could be justified?
If you are working with a single, interval-level variable, then you can treat the cluster memberships as the independent variable in an ANOVA.
Even though theoretically you could get 0 SSE, this is highly unlikely. In general, lower SSE is always better. If you think the SSE is high, try to increase the number of clusters. It could be that the SSE is high because some points are further away and therefore they can be grouped into a separate cluster.
Hi Bryar Hassan,
What algorithm did you apply for cluster analysis?
Sometimes, if some points are very far from each other, there is a suspicion that they are outliers. You could check it by applying the DBSCAN algorithm, or other techniques.
Other times, the clustering algorithms do not normalize the data ([0,1]): you could try to normalize them separately.
Let me know if you need some reference.
Check this:
https://www.researchgate.net/publication/226748490_Clustering_Methods.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs