The Support Vector Machine Based On Uncertainty Variable And Its Application

Support vector machine (SVM) shows many special advantages when processed the classification in real random variable of probability space; however, it is a ticklish question when SVM processed the classification of uncertainty variable which existed in the real life. To tackle the foresaid problem, one support vector machine algorithm which is based on the uncertainty variable is presented. First, we make a method of expert's experimental to collect data, and determine the expert Empirical Membership Function by the Expert's Experimental data. We take the Empirical Membership Function in place of the fuzzy degree of membership which determined by the space position. Simultaneously, to deal with non-equilibrium uncertainty data classification problem, an improved algorithm of unbalanced uncertainty data SVM is constructed with the Fuzzy C-Means Clustering algorithm. Experiments show that the uncertainty support vector machine and unbalanced uncertainty support vector machine perform better in the process of dealing with the uncertainty data.