Thread by @svpino

We usually talk about two main types of machine learning models:

• A Classification model
• A Regression model

They are different, and it's essential to understand why.

↓ 1/6

Whenever the result of your predictions is categorical, you have a classification model.

For example, when your prediction is a binary value (True or False,) or when you want to predict a specific animal from a picture (Lion, Zebra, Horse.)

↓ 2/6

If the result of your predictions is numerical, you have a regression model.

For example, returning a stock's future price, the value of a house, or tomorrow's temperature.

↓ 3/6

Be aware that "Logistic Regression" is a classification model, not a regression one!

I know. Dumb naming. But it is what it is.

↓ 4/6

A few examples of problems so you can determine whether they are classification or regression:

1. Predicting the age of a person
2. Predicting the nationality of a person
3. Predicting whether an email is spam or not
4. Predicting the total revenue from a product

↓ 5/6

5. Predicting the number of products sold
6. Predicting whether it will rain tomorrow
7. Predicting how many inches will snow tomorrow

Answers: RCCRRCR

6/6

Many people have commented (rightfully so) about my statement saying that "Logistic regression is dumb naming."

Logistic regression is a regression algorithm that's used in classification tasks.

So it's the name really dumb that's the case?

This poses a different question: should we classify methods and techniques by how their work internally or by the task they help solve?

I think the correct answer should be both, right?