Algorithm Research on Fuzzy Linear Separable Support Vector Classifier

When the training subset of a support vector machine contains fuzzy information, the support vector machine won´t work. A method for calculating fuzzy linear separable support vector classifier is discussed. With the given confidence level, convert the fuzzy classification problem into finding the fuzzy chance constrained programming, and establish a solution finding theory with fuzzy chance constrained programming rarr crisp equivalence programming rarr dual programming. A algorithm for fuzzy linear separable support vector classifier is derived: given the fuzzy training subset of a fuzzy linear separable problem and selecting appropriate confidential level, construct appropriate quadratic programming equation; find the optimal solution of the quadratic programming; construct optimal separating hyperplane, find the optimal classification function, and establish new training subset; construct support vector regression machine with the new training subset, get the regression functions, and establish the membership function of the optimal classification function from these; lastly, find the optimal solution. The algorithm is also verified with given data.