Hierarchical Clustering - Agglomerative , Divisive

Hello,

I am (name) from LearnVern,

In our previous tutorial of Machine Learning we saw how K means Clustering works.

Now we are proceeding to see our next algorithm that is hierarchical (hi-raar-kikal) clustering.

Here 'hierarchy', (hi-raar-ki) you must have surely heard this word, meaning one at the top, then 2 below it, and furthermore below them two even more of each, so this is known as hierarchy, this is something like tree structure only,

So, let's See this also,

This is also an unsupervised Clustering algorithm, and works upon a tree structure, and on the basis of the tree it makes the grouping of Clusterin

Now, let's see what are the approaches in them.

In this an approach is Agglomerative, meaning of Agglomerative is bottoms up approach,

Meaning it will start from down to up,

In this What happens is that, " each observation starts in its own cluster and pairs of clusters are merged when moves up in the hierarchy", meaning if we have 10 items, then first all of them will become clusters and then we will find some similarities and then on the basis of that will group two or three individually, so in this process hierarchy will move upwards, and with continuous grouping it will finally form itself in a particular cluster only,

So if you want you can cut in between this hierarchy and you will get that much of clusters that you want.

Next, approach is Divisive, meaning top down approach, what happens in this is that, all the items that means all the 10 items in the beginning are in one cluster, next looking at their differences, each of them are split one after another, and this way this top down approach is executed.

So, here we are seeing as to how these clusterings are done,

Now, what is the benefit in clustering?

So in clustering we do not have a requirement of any prior information, and this is easy to implement.

Let me just you using a paint window, as to How a hierarchy is formed and the grouping is made,

So, suppose I have some items like this, so I have three circular items and some square items.

So, if I talk about Bottoms up approach, then what will happen in this is that, all will be in separate clusters, then we will find the similarity, are these two looking similar?

Are these two looking similar, then we will group them, these are not looking similar, further are these two looking similar, then we will connect this, then are they similar, No!, So we connect these two now,

Are these two similar, yes these two are similar then we will connect this also,

So, I am giving you a high level idea, that on the basis of similarities we started from down to up, that is bottom to up, now you will see that all these have come in a similar group, now if you will cut them and so something like this, then you will find each of them is a separate cluster.

So, this is the way we apply the method.

Next, is top to bottom, all are in one clusters at beginning, now we will check the differences and perform the splitting, so by splitting it we will make them as separate clusters,

So, this type of diagram that we make is called a dendrogram, and we cut this dendrogram from the middle, then our clusters are formed.

So, let's go back, so we saw that how basically hierarchical clustering is done,

Ok!

Its application is there in social science, medicine, biology, so at multiple places, wherever we find something new and along with that there is no label, then there we do grouping and apply clustering techniques.

So, friends let's conclude here for today, this session we will stop here, and it's further parts we will cover in next sessions.

So, keep learning and remain motivated.

Thank you.

If you have any queries or comments, click the discussion button below the video and post there. This way, you will be able to connect to fellow learners and discuss the course. Also, Our Team will try to solve your query.