SDS 504: Classification vs Regression — with Jon Krohn

Outcome Types

Understanding the distinction between discrete and continuous outcomes is crucial in differentiating classification from regression problems. Jon Krohn explains that classification problems involve predicting discrete classes, such as determining if a movie review is positive or negative, which can be coded as integers like 0 or 1 1. In contrast, regression problems predict continuous numeric values, such as the future value of a house or stock price 2.

With a regression problem, we would typically use float values, so something with a decimal point, because we have this continuous outcome that we're predicting as opposed to a discrete outcome.

--- Jon Krohn

Both classification and regression are examples of supervised learning problems, where the goal is to predict outcomes based on input data 2.

   

Supervised Learning

In the realm of supervised learning, classification and regression serve as key subcategories, each with distinct input-output structures. Jon Krohn describes supervised learning problems as those where input data, denoted as x, is used to approximate an output, y, which is the desired outcome 3. Classification problems, like determining if a movie review is positive or negative, are examples where the output is a discrete class 4.

Our goal in supervised learning is to have some function that uses the input x to approximate the outcome y.

--- Jon Krohn

Regression problems, on the other hand, involve predicting continuous values, making them distinct yet integral parts of supervised learning 3.