Asymmetric Dual Possibilistic Regression Model by using Pairing nu Support Vector Networks

This research introduces a new and effective asymmetric dual regression model by combining the advantages of possibilistic regression model and paired nu support vector machine (pair-v SVM). Our algorithm is able to best examine the ambiguity in a given data set from internal and external sides. Our algorithm estimates the outer boundary and inner boundary of the uncertain area for the predicted output. Based on the strategy of pair-v SVM, our algorithm find the solutions of four smaller SVM types of quadratic programming problems (QPP) instead of one big QPP to seek the up and down limits of the necessity and possibility model. This scheme greatly speeds up the training speed for our algorithm. Our model adopts the radial kernel, which offers a unified structure for the proposed method, which can handle crisp and vague input at the same time. The experimental results prove the efficiency and effectiveness of our algorithm.