Due to rapid growth in information technologies, millions of data are being generated at every time span. Data storage and processing are more expensive in terms of memory, resources, and time. The traditional machine learning classifier performs well when all the data is available along with the label information at the time of training, which may not be possible in every circumstances. Semi-supervised techniques overcome this limitation. With limited labeled data, semi-supervised learning makes use of both classification and clustering techniques for constructing efficient classifiers, and there is a demand to update the classifier periodically. Unlike traditional approaches, which use the entire data set for updating the classifier, our method updates in an incremental fashion. This paper proposes a framework for incremental classifiers in a semi-supervised environment. Experiments are conducted on benchmark datasets with the proposed approach and compared to the conventional method.