Research Of Credit Evaluation Model On Framework Approach

People's lives has grown closer to credit with the progress of society and the development of the internet increasingly.Due to the decentralization of data and the different standards of evaluation methods,the current credit system need to improve,including the index system of influencing credit and assessment model.This thesis supplements the index system and applies the framework approach to express index system of influencing credit to solve several problem,inluding the backword indexes,the difficult fusion with decentralized data and lack of consideration for special case.Besides,this thesis proposes a credit evaluation scheme based on the framework approach.The indexes influencing credit changes potentially with the progress of the times.There will be more factors will affect the credit.The thesis summarizes the possibility index of the credit system from the aspects of government credit,corporate credit,personal credit and farmer's credit for farmers' group,and uses the framework method to express it.As the advantage of the framework approach,such as clear structure and convenience to extend,the expressed credit system is easy to understand,and using the format like JSON to storage the indicator system will avoid repetitive work produced with data reconstruction when the content of credit system is supplemented.Combined with the training results of the weight of each indicator of the Logistic regression model,this thesis uses the idea of framework approach to improve the original method of matching reasoning.The similarity calculation is used to match the instance,and the property difference is replaced by the probability difference to do weighted mean.Because the data is transformed into probability,the feature item data is transformed into the corresponding probability when the Logistic regression is used to model,which realize the consistency of the data.Comparation with DT and SVM shows that the scheme is feasible in the assessment of the credit,and it improves the accuracy of the assessment in limited data samples.In addition,the scheme greatly reducing the time consumption because it does instance match after the coarse search.