Abstract Data have always played a cardinal role to financial applications, since the power that pumps out of possessing them is translated to safer and more profitable decisions for the corresponding organizations. However, shortage of such kind of data or the inability to access large amounts of them, especially for organizations of smaller range, settles supervised methods as non-auxiliary predictive tools. Thus, techniques that exploit unlabeled data offer the chance of applying even the most advanced strategies for both mining useful patterns from datasets that stem from the collection of sensitive personal information and analyzing their characteristics. In this work, comparisons between several algorithms of Semi-supervised learning and their supervised variants were conducted for examining the applicability of the first category over credit rating problem. The contained datasets come from two different nations (Australia and Germany), concern credit card applications and are publicly available at UCI Machine Learning repository, thus favoring the irreproducibility of our results.