A least squares twin support vector machine method with uncertain data

Twin support vector machine (TWSVM) learns two nonparallel hyperplanes for binary class classification problems. It assumes that the training data can be collected accurately without any uncertain information. However, in practical applications, the data may contain uncertain information. To deal with the uncertain information, this paper puts forward a novel uncertain-data-based least squares twin support vector machine method (ULSTSVM) which is capable of handling the data uncertainty efficiently. Firstly, since the data may contain uncertain information, a noise vector is introduced to model the uncertain information of each example. Secondly, the noise vectors are incorporated into least squares TWSVM. Finally, to solve the derived learning problem, we employ a two-step heuristic framework which trains the least squares TWSVM classifier and updates the noise vectors alternately. The experimental results have shown that ULSTSVM surpasses the existing robust TWSVM methods in training time and meanwhile achieves a better classification accuracy. In sum, ULSTSVM adopts a noise vector to model the uncertain information and transforms the quadratic programming problems of TWSVM into linear equations, which makes us have a better classification accuracy and higher training efficiency.

Automated summary

This paper proposes a novel uncertain-data-based least squares twin support vector machine method (ULSTSVM) that can efficiently handle data with uncertain information. By introducing noise vectors to model the uncertain information and utilizing a two-step heuristic framework, ULSTSVM achieves better classification accuracy and higher training efficiency.