부동산정보 기반 개인 및 개인사업자 대안신용평가모형 개발

Recently, in credit ratings of individual and sole proprietor, alternative credit rating models that utilize various non-financial information in addition to credit information are being actively developed. By using this non-financial information to reflect factors that were not reflected in the existing credit rating model, an individual or sole proprietor with insufficient financial history can make a more accurate evaluation. In particular, there is a growing need for an alternative model that can prevent various real estate accidents that are causing a lot of damage to ordinary citizens, such as real estate lease fraud, which has recently become a social issue. In this study, we propose an alternative credit rating model based on real estate information, such as the number of real estate properties owned, publicly announced and predicted prices, various rights information (provisional registration, temporary seizure, leasehold, mortgage, etc.), and real estate accident information(auction, public auction, lease registration order), targeting individual and sole proprietor who own real estate. The proposed model consists of a real estate model and an owner model as sub-models, and in the case of sole proprietor, a workplace model is added. Factors with low frequency of occurrence but high default rates are used as filtering factors, and if they exceed the filtering factor criteria, a grade suitable for the risk level is given. The model development process was as follows: Based on the candidate factors, univariate significance analysis through primary categorization, significance analysis through secondary categorization, correlation analysis, and multivariate analysis were conducted in order. In multivariate analysis, we applied the logistic regression analysis methodology, which shows high efficiency in model stability and good/bad discrimination ability. As a result of the verification, the stability and discrimination power of the proposed model were very excellent, and it showed a very high discrimination performance improvement effect through combination with the existing model. In particular, it can be seen that the combination of the real estate model (including real estate filtering factors) consisting of only real estate information and the existing model alone maintains very good discrimination power. Therefore, it will be possible to operate a model with high discrimination power even if the sub-model real estate model and real estate filtering factors are applied first before applying the proposed alternative model. In addition, we expect that the proposed alternative credit rating model will enable more detailed and sophisticated marketing and risk management for individual and sole proprietor.