1. Question 1

SVM multi-class classification strategy?

• ❌ Combine supervised + unsupervised

• ❌ One classifier per class

• ❌ Single combined classifier

• ✅ One classifier per pair of classes (One-vs-One)

Explanation:
Binary SVMs → use One-vs-One for multi-class problems.

---

2. Question 2

Why use median at leaf nodes with skewed salary data?

• ❌ Minimizes MSE

• ✅ Reduces impact of extreme values

• ❌ Mean is inaccurate

• ❌ Mean is hard to compute

Explanation:
Median is robust against outliers in skewed data.

---

3. Question 3

Effect of increasing decision tree complexity?

• ✅ Bias decreases, variance increases

• ❌ Bias increases

• ❌ Both constant

• ❌ Both decrease

Explanation:
More complex trees → fit training data better → overfitting.

---

4. Question 4

Finding unusual transactions?

• ❌ Predicting trends

• ✅ Identifying patterns that deviate from normal transactions (Anomaly Detection)

• ❌ Predefined classification

• ❌ Simple grouping

Explanation:
Goal = detect outliers/anomalies.

---

5. Question 5

Productivity increases → slows → stabilizes. Best regression?

• ❌ Exponential

• ❌ Logarithmic

• ✅ Polynomial regression

• ❌ Linear

Explanation:
Nonlinear curve with rise → plateau → polynomial fits well.

---

6. Question 6

Binary classification based on proximity?

• ❌ Logistic regression

• ❌ Decision tree

• ✅ K-nearest neighbors (KNN)

• ❌ SVM

Explanation:
KNN classifies based on nearest neighbors.

---

7. Question 7

Advantage of PCA before clustering?

• ✅ Transforms into high-variance principal axes revealing key features

• ❌ Removes all unimportant features

• ❌ Automatically segments

• ❌ Reduces to one component

Explanation:
PCA keeps maximum variance directions → simplifies clustering.

---

8. Question 8

Faster alternative to gradient descent for large datasets?

• ❌ Backpropagation

• ❌ Grid search

• ❌ Least squares

• ✅ Stochastic Gradient Descent (SGD)

Explanation:
SGD updates weights using small batches → much faster.

---

9. Question 9

Model misclassifies loyal customers as churn risks — fix?

• ❌ PCA

• ❌ Use SVM

• ❌ Add more churn data

• ✅ Adjust the classification threshold

Explanation:
Shifting threshold reduces false positives.

---

10. Question 10

Start with each customer as its own cluster → merge upward?

• ❌ Density-based

• ❌ Divisive

• ✅ Agglomerative clustering

• ❌ Partition-based

Explanation:
Agglomerative = bottom-up clustering.

---

11. Question 11

Why is DBSCAN ideal?

• ❌ Daily travel routines

• ❌ Forecast purchase trends

• ❌ Satellite green cover

• ✅ To isolate rare sensor events in IoT data

Explanation:
DBSCAN excels at detecting outliers/anomalies.

---

12. Question 12

Preserve local AND global structure in high-dimensional data?

• ❌ PCA

• ✅ UMAP

• ❌ Dimensionality reduction not used

• ❌ t-SNE

Explanation:
UMAP preserves both local + global structure better than t-SNE.

---

13. Question 13

Tool for visualizing ML insights?

• ❌ Pandas

• ✅ Matplotlib

• ❌ Scikit-learn

• ❌ NumPy

Explanation:
Matplotlib is the core Python visualization library.

---

14. Question 14

How is ML different from traditional programming?

• ❌ Writes code faster

• ❌ Generates random rules

• ✅ Learns from data to make predictions

• ❌ Hand-coded trees

Explanation:
ML learns patterns instead of using explicit rules.

---

15. Question 15

Library for matrix operations and linear algebra?

• ❌ Scikit-learn

• ❌ Pandas

• ❌ Matplotlib

• ✅ NumPy

Explanation:
NumPy = fast vectorized operations + linear algebra.

---

🧾 Summary Table

---