# Difference of nearest-neighbour clustering and K-nearest neighbor clustering [duplicate]

we're two students working on a seminar paper (topic: Marketing in the Age of Big Data) where we have to conduct a cluster analysis by using nearest neighbour clustering. unfortunately, we cannot differentiate between nearest neighbour clustering and K-nearest neighbours. At first of all we thought that it is the same just called different. After we've read many papers where it is said that KNN is a supervised machine learning algorithm, while our professor said that the nearest neighbour is an unsupervised algorithm we recognised that there must be a difference. There are a lot of different declarations on the internet, why we are confused now.

Hopefully, someone can help us to solve the understanding problems.

Many thanks in advance and many greetings

## 2 Answers

It depends on how you use the nearest neighbors. If all you're doing is finding points that are close to each other and calling them members of a cluster, then this is an unsupervised application. If on the other hand you use the labels of the nearest neighbors to infer something about a given point (either its class or the value of a continuous target), then you are doing supervised nearest neighbors. That said, the latter application is much more common.

• Thank you very much :-) If we understood you right, then the approach (and how to conduct it) is the same, it's just what you're doing with the results of it. In our case, we have to use a Cluster Analysis (k-mean & n-neighbors) for classifying newly incoming customers (e.g. for real-time bidding or dynamic pricing). Jul 26, 2018 at 16:28
• If you're using the nearest neighbors for classification then you're doing supervised learning rather than clustering. Jul 26, 2018 at 16:32

There is neither a "nearest neighbor clustering" nor a "k-nearest neighbor clustering". Can you provide a reference?

1. You probably mean classification and not clustering. There is nearest-neighbor classification, and k-nearest-neighbor classification, where the first simply is the case of k=1.

2. Maybe your professor isn't very well versed here (seems to be marketing, not science) and meant k-means clustering? It's not using the k nearest neighbors... But it has a k in the name, and it seems to be often mislabeled k-nearest-neighbor clustering...