KMeans Clustering Algorithm - 2024 - Machine Learning Projects

Hey guys, So here comes the tenth blog of the Handwritten notes series which we started. We will be talking about KMeans in this blog. I have uploaded my handwritten notes below and tried to explain them in the shortest and best way possible.

The first blog of this series was NumPy Handwritten Notes and the second was Pandas Handwritten Notes. If you haven’t seen those yet, go check them out.

Let’s go through the KMeans notes…

KMeans is an unsupervised learning clustering algorithm that attempts to group similar data together in clusters.
It is widely used in:
Clustering similar documents.
Clustering customers based on features
Market Segmentation
Identifying similar physical groups
Working of KMeans:
Step 1 – Choose k.
Step 2 – Then randomly assign k cluster centers.
Step 3 – Now assign each data point to a cluster center based on the least distance. In this way, the initial guess of clusters is performed.
Step 4 – We will now find the cluster centroids of these clusters and these centroids will be the new cluster centers.
Step 5 – Now we will update the clusters as we did in Step 3 and Step 4 and in this way, we will keep on updating the clusters till the clusters stop changing.
There is no easy way of choosing the best “k” value. One way which gives satisfactory results is the Elbow method.
First of all, it computes the Sum of Squared Errors for some k values like 2,4,6,8, etc.

SSE is defined as the square of the distance between each member of the cluster and cluster centroids.
If you plot k against SSE, you will see the error decreases as k increases. This is because when the number of clusters increases, cluster size decreases, and hence the distortion is also less.
The idea of the “k” method is to choose the “k” where SSE decreases abruptly.
This produces an elbow effect in the graph as you can see in the image above.