Research And Application Of Personal Credit Automatic Evaluation Method

With the rapid development of computer,database and mobile Internet technology,the data volume and data form accumulated in various industries are becoming more and more abundant.Credit reporting is no longer confined to the traditional financial sector.Credit reporting is gradually changing to multiple industries,large flows and small entities.Data on non-financial systems such as sports,social networking,trading and communications are also collected in personal credit reporting.It brings great challenges to traditional credit reporting models and systems.From the model point of view,the integration of a large number of data sets leads to sparse sample indicators,which makes it difficult to guarantee the stability of the model.From the system point of view,the integration of various data forms,flexible configuration of models for different scenarios and concurrent business of a large number of users all put forward higher requirements for the system.From the two aspects of the model and system,this thesis improves the traditional automatic credit evaluation method.In order to effectively solve the problem of data sparsity,users are grouped according to index coverage,and then grouped according to different grouping refinement models.A hybrid model approach based on user grouping is studied in this thesis.Combining with the automated decision-making system,we can flexibly collect and invoke a large number of credit data sets,and support users to customize the whole credit decision-making process without coding.It integrates acquisition,decision-making and monitoring systems.It promotes the whole credit evaluation system to a more accurate and efficient era.In theory,a large number of cases of personal credit products at home and abroad have been used for reference.I study the theory of credit evaluation in similar products.Including credit scoring based on expert rating,based on logistic regression,naive Bayes,decision trees,random forests and other single models,etc.However,these theories show poor stability when the index coverage is sparse.Therefore,on the basis of the traditional credit theory,a new personal credit evaluation method is refined,which is based on the combination model theory of user grouping.In experiment,100,000 sample data of personal credit report were obtained through data acquisition.It tagged positive and negative samples.Based on this data set,it can be split into training set and test set at random.Compared with the traditional credit theory and the combination model theory based on user grouping,the advantages of this theory are verified by experimental data.This theory has a strong universality,can also be extended to consumer finance,supply chain finance,automobile finance and other emerging financial industries.In engineering,in order to realize the automatic evaluation of personal credit,a more intelligent credit decision-making system is needed.The intelligence and automation of the system are mainly embodied in the following aspects:plug and play data source collector,decision logic of all natural language definition,free drag and drop process designer and automated conclusion assessment.It encapsulated the whole process of personal credit assessment.Zero code for decision process.In aspect of application,the combination model theory based on user grouping is applied to this decision-making system in a case-based way to realize automatic scoring.It also passed a complete functional test and performance test.It is proved that the platform has strong stability and high concurrency capability.It provides technical reference for building the credit system of the whole society.