Support Vector Machine (SVM) is a binary classification model. The so-called binary classification model refers to the classification relationship between many features (independent variable X) and another label item (dependent variable Y). For example, there are many characteristics at present, including height, age, education, income, years of education and so on. The dependent variable is smoking or not, and smoking or not only includes two items. Then the research on the relationship between the five characteristic items of smoking classification is called' binary classification model', but in fact there are many categories of label items (dependent variable y), for example, a certain label item y is' cuisine preference', and there are many cuisines in China, including Sichuan cuisine, Shandong cuisine, Guangdong cuisine, Fujian cuisine, Jiangsu cuisine, Zhejiang cuisine, Hunan cuisine and Anhui cuisine.

The commonly used algorithms of machine learning algorithms are decision tree, random forest, Bayesian and so on. These are all very easy to interpret. For example, the decision tree classifies features according to the demarcation point, the random forest is multiple decision tree models, and the Bayesian model is calculated by Bayesian probability principle. Different from the above, the support vector machine model uses operational planning constraints to find the optimal solution, which is a spatial plane. By combining feature items,' smoking' and' non-smoking' can be completely separated. Finding this space plane is the core algorithm principle of support vector machine.

The calculation principle of support vector machine is complicated, but the popular understanding is not complicated. You only need to know that it needs to solve the' space plane', and you can clearly divide different categories of label items (dependent variable y). Similar to other machine learning algorithms, in the construction steps of support vector machine, it is generally necessary to dimension the data, set the ratio of training data to test data, and set relevant parameters for optimization, so as to achieve good performance in training data and test data.

In principle, the support vector machine model is shown in the following figure.

SPSSAU operation is as follows: