Abstract: We proposed an improved least squares twin support vector machine (SMIILSTSVM) incremental learning algorithm based on ShermanMorrison theorem and iterative algorithm. It solved the problem that least squares twin support vector machine (LSTSVM) did not have structural risk minimization and sparsity. The experimental results show that the proposed algorithm has high classification accuracy and high efficiency, and is suitable for noisecontaining crosssample set classification.

Key words: incremental learning, least squares twin support vector machine, sparseness