Personal Credit Assessment Based On Machine Learning Combination Model

As the country's economic strength has increased,people's disposable income has continued to increase,and the credit economy market is constantly evolving.Residents' consumption concept has gradually shifted from saving money to pre-supporting future consumption.Under such a background,China's consumer credit has shown a good upward trend as a whole.Personal credit assessment plays an important and fundamental role in the credit economy market.It promotes the development of the credit economy and stabilizes the economic market.With the development of consumer credit business,the establishment of personal credit system needs to be continuously improved.At this time,it is particularly important to use new technology to promote the development of personal credit evaluation system.The continuous development of the Internet has made people's lives inseparable from the Internet.The arrival of the era of big data tells us that scientific data can objectively reflect our credit level.The information generated by the behavior of everyone in life can be recorded through data.The information on personal credit mainly includes personal basic information,repayment ability and repayment willingness.The basic personal information mainly consists of age,gender and region.It mainly reflects the basic attributes of this person.The repayment ability is mainly composed of characteristics such as assets,wages,and social relations.The willingness to repay mainly examines whether the person has the corresponding event of default or the severity of the event of default,so the individual The credit rating system consists of these three factors and contains all the data about a person's credit default.The problem of personal credit assessment is a classification problem.The research on personal credit problem has become more mature by using statistical methods.When researching personal credit problems,applying machine learning technology to it can improve a lot of efficiency and a large amount of data.The process also plays an important role.In the machine learning,various classification algorithms can deal with the personal credit evaluation model.In view of the fact that scholars have perfected the research on a single personal credit evaluation model,this paper studies the advantages and disadvantages of the single model and the combined model.The data in this paper comes from the personal credit data provided by a foreign credit institution on the Kaggle data platform.The amount of data reaches 150,000,covering all the characteristics of the personal credit system construction,and is the data that can objectively reflect a person's credit score.The article first preprocesses the data,combines the deletion method and the padding method,processes the missing values in the data,and then deletes the outliers in the data to obtain a complete data set.Then three-quarters of the data set is used as the training set to learn the model,and one quarter is used as the test set to evaluate the pros and cons of the model,to ensure that the data set of each model is the same,and the data set of each model is tested the same.The results of comparing the generalization capabilities of each model are scientific.In this paper,we choose the decision tree algorithm and the logistic regression algorithm to construct two single models,and then compare the accuracy between the single models.Firstly,the basic principles of the two algorithms are studied.Then,each algorithm is used to learn the personal credit evaluation model in the complete data set.The two single models are tested on the same test set,and then the classification evaluation results of the two single models are compared.It is found that the logistic regression model is superior to the decision tree model.The combined model has two main methods in the field of personal credit evaluation: serial structure combination and parallel structure combination.The paper firstly combines the tree strategy and the logistic regression algorithm,and uses the output of the decision tree as an explanatory variable,adds it to the characteristics of the logistic regression model,outputs a serial combination model,and then uses the test set to The model test yields the classification evaluation index.It is found that the classification effect of the serial combination model is better than the two single models.After obtaining the serial combination model,the two single models are combined in parallel.The parallel combination model allows the output results of the two models to be combined into a new output according to different weights,and the appropriate weights are selected to construct an optimal combination model.Then the constructed combined model is tested on the test set,and the classification results are obtained.It is found that the optimal combination model has better classification effect than the serial combination model.The results of this study show that the prediction accuracy of the single decision tree model for the personal credit assessment problem is worse than the logistic regression model.Compared with the single model,the combination model is either a serial model or a parallel combination model for personal credit.The evaluation research should be better,the prediction accuracy should be higher,and the classification effect is more accurate and more explanatory in the field of personal credit evaluation,which obviously reduces the credit risk.In the combined model constructed by decision tree and logistic regression,the classification effect of the optimal parallel combination model is better than the serial combination model.For the future personal credit evaluation research,in the field of combined model research,many methods can be used to study the combination of different structures based on different single models,and there is still a large research space,which is the research of future personal credit evaluation model direction.