A twin support vector machines(TwinSVM) performs better than a support vector machine(SVM) in dealing with the class-imbalanced classification problem. However, the TwinSVM has weak generalization abilities when there are not enough training data. Therefore, a knowledge embedded transfer twin support vector machine(KE-T-TwinSVM) is proposed. The KE-T-TwinSVM inherits the characteristic of the TwinSVM and fully considers the data of target domain as well as the knowledge of source domain by leveraging the knowledge of source domain in the target domain. The experimental results show that the proposed KE-T-TwinSVM has better performance compared with the traditional classifiers in the situation of insufficient data and class-imbalanced data.