What
The classic binary classification algorithm, can solve the multi-classification problems
-
Sigmoid function
official:
Argument value is an arbitrary real number, the range [0, 1] 
Integrate the results:
By the input value to the probability of conversion, it can be used to complete the classification task
Likelihood function:
Log-likelihood:
Why
Simple structure. First with logistic regression test, if not resolved to consider other algorithms