Unsupervised Learning

Question 1

• Supervised vs unsupervised learning

Answer 1

Supervised uses labeled data, unsupervised uses unlabeled data

Question 2

• How unsupervised learning can be helpful during exploration of a dataset

Answer 2

helps explore and understand the underlying structure of a dataset,
making it easier to prepare data for further analysis and model building.

Question 3

• Define a cluster and its attributes

Answer 3

• Groups of similar data points
• Attributes: number, membership, centroids, intertia

Question 4

• Why is it important to standardize data before performing clustering?

Answer 4

ensuring that all features contribute equally, improving the performance and reliability of the clustering algorithm, and preventing biases from features with larger scales.

Question 5

• Are all clusters meaningful? If not, what are the pitfalls and how can we avoid them?

Answer 5

No
pitfalls include noise and outliers, incorrect number of clusters
use elbow plot and silhouette score

Question 6

• Explain the concept of an anomaly and describe how unsupervised learning can help us to detect them.

Answer 6

data that deviates from the normal pattern in a dataset.
Unsupervised learning helps detect them through clustering, density-based, distance-based methods

Question 7

• Describe some of the important roles of a human operator when undertaking unsupervised learning.

Answer 7

• Data cleaning, standardization
• Feature selection
• Parameter tuning
• Engineering judgement and knowledge

Question 8

• Hierarchical vs flat clustering

Answer 8

• Hierarchical - useful for capturing nested structures and provides visual representation through dendrograms. It is computationally intensive and less scalable.
• Flat - Efficient and scalable, requires the number of clusters to be predefined, and is suitable for flat, non-hierarchical data structures.