Course contentsShow
← Machine Learning and Deep Learning
Lesson 123 of 3,5382. Machine Learning Mindset and Core ConceptsFree lesson

The Importance of Problem Formulation

How framing a real-world problem as an ML task determines what's possible to achieve.

The Importance of Problem Formulation

What you'll learn: How the way you frame a real-world problem as an ML task fundamentally determines what solutions are possible and what success looks like.

The Frame Defines the Game

Before writing any code or choosing algorithms, you must translate a messy real-world situation into a well-defined ML problem. This problem formulation step is where you decide:

  • What exactly are you predicting or learning?
  • What counts as "input" (features)?
  • What counts as "output" (target)?
  • What does success look like (how will you measure it)?

Get this wrong, and even the most sophisticated model will solve the wrong problem beautifully.

Why It Matters

Think of problem formulation like choosing the rules of a game before playing. If you want to help doctors diagnose diseases, you could frame it as:

  • Classification: "Does this patient have disease X? (yes/no)"
  • Ranking: "Which diseases are most likely, in order?"
  • Severity prediction: "How advanced is the disease? (scale 1-5)"

Each framing leads to different data needs, different models, and different notions of what "good performance" means. A model that's 95% accurate at yes/no diagnosis might be useless at ranking diseases by likelihood.

The formulation also determines what's possible. If you frame a problem requiring future prediction but only provide historical data with no time component, you've made success impossible from the start.

The Constraint on Everything

Your problem formulation constrains everything downstream: data collection, feature engineering, model selection, evaluation metrics, and deployment strategy. A poorly formulated problem wastes resources solving something nobody needs.

Key Takeaway: Problem formulation is the foundational step that translates real-world goals into a learnable ML task—choose it carefully, because it determines what's achievable and how you'll measure success.