SUPPORT VECTOR MACHINES

 

In this chapter, we discuss a relatively new regression analysis technique called support vector machines, or SVM for short. SVM is considered one of the best classifiers in supervised learning for analyzing complex data and downplaying the influence of outliers.

Developed within the computer science community in the 1990s, SVM was initially designed for predicting numeric and categorical outcomes as a double-barrel prediction technique. Today, SVM is mostly used as a classification technique for predicting categorical outcomes—similar to logistic regression.

 

Figure 27: Logistic regression versus SVM

 

In binary prediction scenarios, SVM mirrors logistic regression as it attempts to separate classes based on the mathematical relationship between variables. Unlike logistic regression, however, SVM attempts to separate data classes from a position of maximum distance between itself and the partitioned data points. Its key feature...