The paper focuses on the construction of support vector machines when the outputs of training points are triangle fuzzy numbers. First we transform the fuzzy classification problem to chance constrained programming with fuzzy decision, and solve this programming using fuzzy simulation based evolutionary algorithm. Based on this, we proposed the fuzzy linear support vector machines (FLSVM). It can deal with classification problems with fuzzy information as well. At the end, the definition of fuzzy support vector set is given, which can declare the characters of FLSVM.
