1. Question 1

Why run an unsupervised model first with no diagnosis labels?

• ❌ Rank patients by admission time

• ❌ Compress features into one index

• ❌ Assign diagnosis codes automatically

• ✅ To uncover natural patient groups that share similar vital sign patterns

Explanation:
Unsupervised learning finds hidden structures when no labels exist — perfect for grouping patients by similarity.

---

2. Question 2

How does hierarchical clustering help determine the number of clusters?

• ✅ Visualizes similarity levels using a dendrogram to help choose optimal cluster count

• ❌ Produces same clusters every time

• ❌ Automatically removes outliers

• ❌ Designed specifically for high-dimensional data

Explanation:
A dendrogram shows where clusters merge, helping select the right number of groups.

---

3. Question 3

What does a K-means centroid represent?

• ❌ Range of spending

• ❌ Distance between customer and center

• ❌ Number of customers improving

• ✅ The coordinate pair representing the average spend and purchase frequency

Explanation:
A centroid = mean position of all points in a cluster.

---

4. Question 4

Why is DBSCAN good for detecting unusual social media activity?

• ❌ Need exact number of clusters

• ✅ It finds clusters of any shape and detects outliers

• ❌ Assigns every user to a cluster

• ❌ Minimizes user-center distances

Explanation:
DBSCAN naturally identifies dense groups and marks isolated points as anomalies.

---

5. Question 5

Why use t-SNE for 2D scatterplots of customer behavior?

• ❌ Label each segment automatically

• ❌ Enforce linear projections

• ❌ Equal distance between points

• ✅ Maintain neighborhood similarities for visual discovery

Explanation:
t-SNE preserves local structure, making clusters visually meaningful.

---

6. Question 6

Why is PCA suitable for environmental research?

• ❌ Finds nonlinear relationships

• ❌ Combines all variables into one

• ❌ Forces uncorrelated features

• ✅ Reduces data to key components for simpler analysis

Explanation:
PCA highlights the most important variance directions, simplifying analysis with fewer features.

---

7. Question 7

Advantage of using t-SNE for app user interaction visualization?

• ❌ Eliminates noise

• ❌ Guarantees linear transformation

• ✅ Allows similar users to form clusters in low-dimensional space

• ❌ Reduces to one dimension

Explanation:
t-SNE produces meaningful cluster structures for visualizing high-dimensional behavioral patterns.

---

🧾 Summary Table

---