Personal Credit Risk Assessment Based On Feature Space Division And Ensemble Model

The most significant factor impacting an assessment of individual credit default risk is personal credit.The possibility of personal credit risk default increases with rising demand for personal borrowing,leading the entire financial system to become unstable.Therefore,an exhaustive assessment of credit risk is an essential part of the financial system.Traditional credit risk assessment techniques rely overly on personal credit,which is sufficient to fulfill the big data era’s present needs for timeliness,accuracy,and diversity of data.current high dimensionality and sparseness of data make feature selection extremely important,which makes it difficult for traditional credit risk models to make accurate assessments.Meanwhile,default samples in financial data are sometimes quite small,making it difficult for traditional credit risk models to make accurate assessments.The following three topics are clearly listed in the research for this paper:1.A feature space division based on a credit risk assessment.One person’s credit risk assessment can improve the default risk as the feature selection process hired in traditional credit risk assessment is fairly simple.In this research,we present a novel feature space division method that eliminates a subset of features with high relevance and low redundancy through the use of random forest and self-encoder.2.A learning method based on credit imbalance data samples.Random sampling of a few classes of samples is a common method to solve the imbalance of data samples,but the random sampling process has the problems of inaccurate sampling and sampling distortion,which will affect the evaluation of the model.For unbalanced credit data,this paper proposes an ADASYN sampling method based on K-means clustering improvement,which improves the quality of generating new samples,and applies the method to the learning of unbalanced data samples to improve the diversity of minority class samples.3.Individual credit risk assessment method based on integrated learning.Single classifier is a renowned financial risk assessment model that focuses on feature selection,whilst integrated classifier exceeds single classifier when it comes to of generalization performance.For the purpose to precisely record the potential information in the base classifier and enhance the classifier’s accuracy,an integrated learning model was selected in the present research.