How is "Demand Prediction" model not a Linear Regression model?

In this lecture :

Ing suggests that the Demand Prediction model is a logistic regression model. As far as I’ve understood, Log Reg is used for classification, while Lin Reg is used for getting precise values of whatever we are trying to find out, which in this case is the exact probability of whether the clothing item would sell or not.

According to me, it would’ve been a Log Reg if we were trying to just get whether the shirt will “sell” or “not” as we’d be classifying our data into the two groups of “sell” and “not”, but here we want an exact value of probability hence it seems a Lin Reg model. Where am I wrong?


A logistic regression produces a continuous output bounded between 0 and 1. We say the value means the probability of being True. We convert the probability into True / False label by a threshold value. Usually the threshold value is 0.5 such that when the probability of being True is larger than the threshold of 0.5, we say the prediction is True. So,

Logistic regression: bounded, continuous probability output between 0 and 1 → > threshold ? → class True (1) or class False (0)
Linear regression: unbounded, continuous output.

Do you see the difference between mine and yours?



Hi @Mohd_Farhan_Hassan I just want to add this picture to illustrate the explanation of @rmwkwok

Overall, logistic regression is bounded between 0 and 1 while linear regression is a continuos output and is not bounded.

I hope this helps