A novel classification technique based on progressive transductive SVM learning

• A novel classification technique based on transductive SVM learning is presented.
• The paper pointed out the drawbacks of the existing literature methods.
• To mitigate drawbacks, here we exploit SVM, k-nn classifiers and cluster assumption.
• Results were compared with two other TSVM methods existing in the literature.
• Experimental results confirmed the effectiveness of the proposed technique.

The existing semisupervised techniques based on progressive transductive support vector machine (PTSVM) iteratively select transductive samples that are closest to the SVM margin bounds. This may result in selecting wrong patterns (i.e., patterns that when included in the semisupervised learning can be associated with a wrong label) as transductive samples, especially when poor initial training sets are available or when available training samples are biased. To mitigate this problem, the proposed approach considers the distance from SVM margin bounds, the properties of the k-nearest neighbors approach, and the cluster assumption in the kernel space. To assess the effectiveness of the proposed method, we compared it with other PTSVM methods existing in the literature by using a toy data set and six real data sets. Experimental results confirmed the effectiveness of the proposed technique.