Logistic regression is another supervised learning technique, which is basically a probabilistic classification model. It is mainly used in predicting a binary predictor, such as whether a customer is going to churn or if a credit card transaction is fraudulent.
Logistic regression uses logistics. A logistic function is a very useful function that can take any value from a negative infinity to a positive infinity, and output values from 0
to 1
. Hence, it is interpretable as a probability. The following is the logistic function that generates predicted values from 0
to 1
based on the dependent x
variable:
Here, x will be the independent variable and F(x) will be the dependent variable.
If you try to plot the logistic function from a negative infinity to a positive infinity, then you'll get the following S shaped graph:
Logistic regression can be applied in the following scenarios:
Deriving a propensity score for a customer in a retail store of buying a new product that has...