We use Logistic Regression (LR) to solve binary or 2-class classification problems. LR produces a single value between 0 to 1.

In order to find four different categories with the help of LR, we need to create four different models (i.e. One-vs-All technique). One model tries to predict Class 1 vs [Class 2, Class 3, Class 4]. Second model finds Class 2 vs [Class 1, Class 3, Class 4]. Third model is for Class 3 vs [Class 1, Class 2, Class 4] and the last model is to predict Class 4. Once we train four models, we can find the prediction result from the models with the highest score.