SHORT QUESTIONS FOR HIERARCHICAL CLUSTERING

General Introduction

Hierarchical clustering is a way for computers to group similar things together without being told how. It builds a tree of clusters that shows how the groups are formed. This can be shown with a dendrogram (a tree-like diagram).

There are two main types:

• Agglomerative (Bottom-Up): Start with each item in its own group, then slowly combine the closest ones.

• Divisive (Top-Down): Start with everything in one big group, then slowly split it into smaller ones.

Q1. When should you use hierarchical clustering, and when not?

Use hierarchical clustering when:

· You want to understand groupings or structure in data.

· You're working with small to medium-sized datasets.

· You need to visualize clustering with a dendrogram.

Avoid it when:

· You're dealing with very large datasets (due to performance issues).

· You need real-time or very fast clustering.

· You have high noise or irrelevant features (it can skew the results).

Q2. What are the advantages and disadvantages of hierarchical clustering compared to other methods like K-means?

 Advantages of Hierarchical Clustering:

• You don’t need to choose the number of clusters in advance.

• It shows how data groups step by step, giving more insight.

• You can use a dendrogram (tree diagram) to see the clustering clearly.

Disadvantages:

• Slow with large datasets (takes a lot of time and memory).

• Can be affected by noise or outliers.

• Once it makes a decision, it can’t go back (not flexible).

Compared to K-means:

• K-means needs you to pick the number of clusters first; hierarchical doesn’t.

• K-means is faster and works better for large datasets.

• Hierarchical works better when clusters are nested or not clearly separated.

Q3. What are the two main types of hierarchical clustering, and how do they compare?

Two Main Types of Hierarchical Clustering:

A. Agglomerative (Bottom-Up):

• Starts with each point as its own cluster.

• Joins the closest clusters step by step.

• Ends with one big cluster containing all points.

• Most common and easier to use.

B. Divisive (Top-Down):

• Starts with all points in one big cluster.

• Splits off the most different points into smaller clusters.

• Keeps splitting until each point is alone.

• Less common because it’s more complex.

Easy Comparison:

• Agglomerative is like building a puzzle piece by piece.

• Divisive is like taking apart a completed picture.

Q4. What kinds of data or problems are best suited for hierarchical clustering?

When Hierarchical Clustering Works Best:

Hierarchical clustering works well when your data is organized in layers or steps, like a tree.

If your data has levels — such as family trees, topic categories, or groups — it helps you find hidden patterns.

It’s best for exploring data to understand it better, not just for quick results.

Examples:

• Grouping genes that act alike.

• Finding topics within documents (like politics → elections → candidates).

• Seeing if customer groups split into smaller groups with similar behaviors

 Q5. How does the choice of distance metric and linkage method influence hierarchical clustering results, and which linkage method is best?

Ans. The way you measure distance and link clusters in hierarchical clustering changes your results a lot. Different mixes, like Euclidean distance with complete linkage or cosine distance with average linkage, can group data very differently.

Each linking method works in its own way. Single linkage can make long, stretched-out clusters, while complete linkage makes tighter, rounder groups. Ward’s method is good for balanced groups, especially with data that changes smoothly.

There isn’t one best method for all cases—it depends on your data and what you want. You need to find a balance between how clear, accurate, and fast you want the results to be.