A SVM-Based Text Classification Method with SSK-Means Clustering Algorithm

SVM-based classification needs lots of labeled data to train classifier model, but labeling training dataset is a time-wasting and energy-wasting task. Furthermore, the feature space is sparse commonly because of text´s high dimension. All of the factors above can influence the performance of classification. We propose a SVM-based text classification with SSK-means clustering algorithm where little labeled training data are needed. In this approach, training data, including both labeled and unlabeled data, are first clustered with guidance of the labeled data. The unlabeled data samples are then labeled based on the clusters obtained. SVM classifiers can be trained with the expanded training dataset. When the training dataset has only a little labeled data, this method has better performance than SVM classifiers.