Adversarial Semi-supervised Learning for Corporate Credit Ratings

Corporate credit rating is an analysis of risks withina corporation, which plays a vital role during the management financial risk. Traditionally, assessment process based on historical profile corporation usually expensive and complicated, often takes months. Therefore, most corporations, duetothelack in money time, can’t get their own level. However, we believe that although these corporations haven’t levels (unlabeled data), this big data contains useful knowledgeto improve system. In work, its major challenge lies how to effectively learn knowledge from unlabeled help performance Specifically, consider problem adversarial semi-supervised learning (ASSL) for corporate has been rarely researched before. A novel framework (ASSL4CCR) includes two phases proposed address problems. first phase, train normal system via machine-learning algorithm give pseudo Then second applied uniting labeled pseudo-labeleddatato build final model. To demonstrate effectiveness ASSL4CCR, conduct extensive experiments Chinese public-listed dataset, proves ASSL4CCR outperforms state-of-the-art methods consistently.