A New Classification Based on the Method of Similitude Compress Convex Hull

摘要

In this paper, we validate the feasibility of similitude compress convex hull, and design a new classification based on the method. And, we give a judgment about two convex hulls' intersection and other parameters' estimation. Then we deduce the geometrical bisection method based on similitude compress convex hull. Besides, we present another method for compressing convex hull, and the linear separable SVM based on similitude compressing all points. At last, we give the result of numerical simulation about Iris flower data, shows the high time efficiency and accuracy by the algorithm.