Hello, I applied Ward Hierarchical cluster analysis and I used some indexes for the validation of the clusters. Please check the Silhouette score for the 2 and 3 clusters in all months. I want to know that in this case how many clusters we consider, 2 or 3?
Silhouette score, Dendrogram and maps for 2 and 3 clusters are attached herewith.
Thank you.
Asfandiyar Khan Talking about the Silhouette score, the best value is one, and the worst value is -1. Values near 0 indicate overlapping clusters. As I can see in your Silhouette plot, you have a higher value with c=2 so I would go with it. Also, because you have a dendrogram, You can check your dendrogram and you would see that you have a higher distance between 2 clusters than between 3 clusters.
But also would be interesting to see how you did these graphs and what data you used for them? Do you cluster your data separate for each month?
@Mantas Lukauskas Sir, thank you for the answer. The dataset was Tropical Rainfall Measuring Mission (TRMM) satellite data and its temporal resolution was daily which was converted to monthly for monthly analysis. It was obtained for each grid point covering the study region. The dendrogram was prepared in jupyter lab for the month of January. The clustering was performed for each month and Silhouette values are shown. Thank you
On how many clusters for analysis, I suggest you follow Article A Comprehensive Survey of Clustering Algorithms
andArticle A Survey on clustering Current status and challenging issues
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts