-It is a supervised type of machine learning algorithm.
-It is used for the classification problems and not for regression problems, still the name given is logistic regression.
-Because at the back it also uses linear model or we can also say that classification method uses the same concept as linear regeression. That is the reason it is called as logistic regression.
-And the word logistic regresent here, the logloss function.
-Target variable should be in the categorical form.
It can be two or more categories.
- Binary class classification: >> It does have TWO Categories
(spam/ not-spam) (Yes/No)
- Multiclass classification: >> It does have MORE-THAN-TWO Categories.
low/medium/high -->> Ordinal (there is a Relation between them)
mango/apple/banana -->> Multinominal (there is No-Relation between them)
It simply tries to convert Independent variable into an expression of probabilities that ranges in between 0 and 1 w.r.t Dependent variable.
With the help of Sigmoid Function it converts the liner line to S-shaped curve which is called as Sigmoid Curve.
1/1+e^(-y)
where,
y = equation of line
e = euler's constant = 2.718