I'm having a dataset which consists of 2500+ houses information like price, sqft area, etc. I wanted to group similar houses on the basis of the house's location, price, number of bedrooms, and bathrooms. So taking these 4 as the input parameters, I applied k-means clustering and for determining the number of clusters (i.e. the value of k) I used the Silhouette analysis technique and get k = 208 as the one with best silhouette score so the divided my dataset into 208 clusters with K-means clustering. Now I created a single sample data having random location, price, number of bedrooms, and bathrooms and predicted the cluster-number this sample belongs to and analyzed the data points of that cluster. My problem is that I also want to analyze the data points of the most similar cluster(s) to the single sample instance I created. How can we do this?
Parva Jain
If I understand you correctly, then here is my suggestion:
You can merge the 2 datasets and run a clusterer method on the resulting data. Then, explore the instances of each cluster to see which data they originally belong to.
HTH.
Dr. Samer Sarsam
K-means will generate the central coordinates of each cluster, and you can calculate the distance between the sample and each cluster.
Once clusters have been identified, you would able to find the corresponding sample's cluster! right.
To find the top m neighbors' data points within the cluster, you can set a marginal distance. It should be less than your minimum cluster intra-similarity (less than farthest data points coordinates). Then design a plain local search based on the aforementioned criteria and find the distance between central coordinates and data points caught by local search. Finally, order them descendingly and select the top m points WRT the desired search intensity.
You can find the inter-cluster distances using any distance formula. The ones which provide the smallest value(s) are the closer ones. Similarly the other way.
- Badges
- Science topic
- Similar topics
- Comment