Unsupervised Learning

• Clustering (Divide by similarity)

Models are trained on unlabeled data and must discover structure on their own.

• We don't have the right answer for the data set.

Customer Segmentation

• X:
  • Purchase frequency
  • Average order value
  • Product categories
  • Visit patterns
• y:
  • Groups such as "high-value", "occasional", "price-sensitive"

Email categorization

• X: Take a bunch of emails and create groups of messages. We can later name those groups (e.g. work, friends, family, marketing, spam, etc.)

• News items

• DNA sequences how much certains Genes are expressed

• Social network analysis

• Market segmentations

• Astronomical data analysis

• Cocktail party algorithm (separating two voices as recorded by two microphones) (Noise cancellation)