Laplacian support vector machines for medical diagnosis

A semi-supervised learning method is presented for medical diagnosis owing to the large amount of unlabeled samples of training model. Laplacian graph which is state-of-the-art method in manifold regularization is used to smooth the probability density functions. The Laplacian regularization term is added to SVM algorithm constituted LapSVM which would be applied to medical data classification and verified on Breast Cancer Dataset, Mammographic Mass Dataset and Thyroid Gland Dataset. Experiments indicate that LapSVM can achieve a better performance using the small labeled samples and large unlabeled samples.