Last article ：「14」 Support vector machine —— I finished , Who supports ？ Who is against ？, We go through SVM This is the most common machine learning algorithm . In this article, we use a very simple python Let's practice the actual project SVM And deepen understanding of .

review

SVM It's a two category model , The data processed can be divided into three categories ：

1. Linearly separable , Maximize... By hard spacing , Learning linear classifiers
2. Approximately linearly separable , Maximize... By soft spacing , Learning linear classifiers
3. Linearly indivisible , By kernel function and soft interval maximization , Learning nonlinear classifiers

Linear classifier , It corresponds to a straight line on a plane ; Nonlinear classifier , It corresponds to a curve on a plane .

Hard intervals correspond to linearly separable data sets , All samples can be classified correctly , That's why , Affected by the noise samples , Not recommended .

The soft interval corresponds to the usual data set （ Approximately linearly separable or linearly nonseparable ）, Some samples near the hyperplane are allowed to be misclassified , This improves the generalization performance .

Here's the picture ：

The solid line is obtained by maximizing the hard interval ࿰