As data volumes grow exponentially, manually labeling all data becomes impractical, driving the development of Domain Adaptation techniques. Domain Adaptation allows models to transfer knowledge from a source domain to a similar target domain with different feature spaces, addressing the challenge of domain shifts caused by varying data collection methods. This paper explores several unsupervised domain adaptation approaches, including Transfer Component Analysis (TCA) and Geodesic Flow Kernel (GFK), and evaluates their performance improvements when augmented with a minimally semi-supervised target domain containing only 3–5% labeled data. Our findings highlight significant performance gains and underscore the importance of semi-supervised learning in enhancing domain adaptation effectiveness.