1.2.2. Matching ML Techniques to Use Cases (Regression, Classification, Clustering)

First Principle: The nature of the business question dictates the required ML technique. Predicting "how much" is regression, "which one" is classification, and "what groups" is clustering.

This is the most fundamental mapping an AI Practitioner must make.

• Use Case for Regression (Predicting a number):
  • Question: "How much will this house sell for?"
  • Question: "What will our revenue be next quarter?"
  • Question: "How many days until this customer's subscription expires?"
• Use Case for Classification (Predicting a category):
  • Question: "Is this email spam or not spam?"
  • Question: "Will this customer churn (yes/no)?"
  • Question: "Which of these three marketing campaigns will this user respond to?"
• Use Case for Clustering (Finding groups):
  • Question: "What are the natural groupings of our customers based on their behavior?"
  • Question: "Can we group these news articles by their underlying topic?"
  • Question: "Are there distinct patterns of network activity that we can group together?"

Scenario: You are in a planning meeting. A colleague wants to know "which of our customers are most similar to each other," and another wants to know "what the estimated lifetime value of this specific customer is."

Reflection Question: Which ML technique (regression, classification, or clustering) would you propose for each question, and why?

💡 Tip: Focus on the desired output. If it's a number, think regression. If it's a category or a label, think classification. If it's about discovering groups with no predefined labels, think clustering.