[02468] Intuitionistic fuzzy proximal twin svm with fuzzy hyperplane
Session Time & Room : 3C (Aug.23, 13:20-15:00) @D501
Type : Contributed Talk
Abstract : Twin support vector machine (TWSVM) is a contemporary machine learning technique for classification and regression problems. However, TWSVM is sensitive to noises as it ignores the positioning of the input data samples and hence fails to distinguish between support vector and noises. To overcome this issue, we propose a novel Intuitionistic fuzzy proximal twin svm with fuzzy hyperplane (IFTPSVM-FH). Instead of addressing two quadratic programming problems like in TWSVM, two non-parallel classifiers are obtained by solving two systems of linear equations which makes the model more efficient. The two major features of the proposed approach are that it gives an intuitionistic fuzzy number based on the relevance to each data vector and that the parameters for the hyperplane, such as the components of the normal vector and the bias term, are fuzzified variables. With the use of fuzzy variables, the proposed fuzzy hyperplane effectively captures the ambiguous character of real-world categorization tasks by representing vagueness in the training data. The proposed approach uses local neighbourhood information among the data points and also uses both membership and non-membership weights to reduce the effect of noise and outliers. By incorporating nonlinear kernel functions into the feature space, the method can be used to detect complex patterns or non-linearity in the dataset. We have applied our method on real-world classification tasks and concluded that it performs incredibly well in comparison to other approaches. In order to demonstrate the practical application of the proposed model, we use it for the predict the trends of the stock market.
