One of the first concepts in machine learning is understanding the difference between classification and regression. Although both use data to make predictions, they answer different kinds of questions.
Classification predicts a category or label. The answer belongs to a group, such as:
- Will a patient be readmitted? Yes or No
- Is a claim potentially fraudulent? Fraud or Not Fraud
- Does a patient have diabetes documented? Present or Not Present
- Is a member likely to qualify for a care management program? Eligible or Not Eligible
Regression predicts a continuous numeric value. Examples include:
- What will the patient’s total annual healthcare cost be?
- What is the expected length of hospital stay?
- What is the projected monthly PMPM (Per Member Per Month) cost?
- How many emergency department visits might occur next quarter?
A simple way I remember the difference is:
- Classification = Choose a category.
- Regression = Predict a number.
Healthcare Example
Suppose a healthcare organization wants to improve chronic disease management.
- If the goal is to predict whether a patient is likely to be readmitted, that is a classification problem.
- If the goal is to estimate how much that patient’s healthcare costs will be over the next year, that is a regression problem.
Both approaches are widely used in healthcare analytics. The choice depends on the business question we are trying to answer.
🌱 Study Note: Before choosing a model, ask: “Am I predicting a group or a value?”