## K-Means,AggloN Custering Methods

What next after creating these custers?

What is the use of these Custers?

How can i get the prediction for a particular data?

What is the use of these Custers?

How can i get the prediction for a particular data?

- Bhagwant
**Posts:**10**Joined:**Tue Feb 14, 2012 3:17 pm

K-means and AggloN algorithms collectively known as Clustering, which are part of the Unsupervised Learning set of machine learning techniques.

Both will allow to automatically group different elements of an item set based on certain characteristics, and identify "clusters" of related items.

In the case of k-means, for example, you pass a set of items with a number of features (or "dimensions") and a set of tentative centroids (usually randomly selected to match some of the items). After running several iterations, the algorithm will converge to a local minimum, which represents the positions of these centroids that minimize the distance to the items in the set.

Since the convergence is to a local minimum and selecting the number of initial centroids can be tricky, it's not uncommon to execute several runs with different number of centroids and random startup locations, and calculate the convergence using certain measurements (F-measure, Dunn Index, etc.), using the "elbow rule" to identify the most efficient number of starting centroids.

Agglomerative clustering (AggloN) uses a hierarchical agglomerative model but it's otherwise similar in principle.

Flavio

Both will allow to automatically group different elements of an item set based on certain characteristics, and identify "clusters" of related items.

In the case of k-means, for example, you pass a set of items with a number of features (or "dimensions") and a set of tentative centroids (usually randomly selected to match some of the items). After running several iterations, the algorithm will converge to a local minimum, which represents the positions of these centroids that minimize the distance to the items in the set.

Since the convergence is to a local minimum and selecting the number of initial centroids can be tricky, it's not uncommon to execute several runs with different number of centroids and random startup locations, and calculate the convergence using certain measurements (F-measure, Dunn Index, etc.), using the "elbow rule" to identify the most efficient number of starting centroids.

Agglomerative clustering (AggloN) uses a hierarchical agglomerative model but it's otherwise similar in principle.

Flavio

- flavio
- Community Advisory Board Member
**Posts:**73**Joined:**Wed Apr 27, 2011 8:59 pm

2 posts
• Page

**1**of**1**### Who is online

Users browsing this forum: No registered users and 1 guest