As explained in the previous update, I am going to tell about the pros and cons of k-means, k-mediods and DBSCAN clustering methods in this update.
K-Means:
K-Means is like a quick organizer that has its ups and downs. On the positive side, it’s super fast and easy to use. Think of it as sorting different-colored marbles into boxes quickly. But here’s the catch: it might give you different results depending on where you start, and it only works well when the groups are round like circles. Also, you have to tell it how many boxes (groups) you want before it starts sorting.
K-Medoids:
K-Medoids is like a careful organizer. It’s better at handling messy situations where you have some strange items in your data. Imagine sorting rocks of different sizes. Instead of just using the center of each group, it picks one rock from each group as a representative. It’s not as speedy as K-Means, but it doesn’t get confused by odd items. However, like K-Means, you still need to say how many groups you want from the beginning.
DBSCAN:
DBSCAN is like a detective. It can figure out how many and what kind of groups are hiding in your data all by itself, which is pretty cool. It’s great at finding groups of different shapes, like stars and clouds. But here’s the thing: you need to give it some hints. You have to tell it how close things should be to belong to the same group and how many things make a group. Sometimes, DBSCAN can struggle when groups have different levels of crowdedness, and it might not be 100% sure about things near the edges of groups.
In summary, K-Means is a fast sorter but assumes round clusters, K-Medoids is careful with messy data and outliers, and DBSCAN is a detective that can find clusters of different shapes but needs some direction.