Application Of Random Forest Method In Personal Credit Risk Analysis

With the continuous reform of the financial services industry,the transformation of consumer attitudes and the rise of Internet finance,the personal credit business has developed rapidly and its scale has continued to expand,but the credit risk has also increased.Therefore,strengthening the management of personal credit risk and finding a scientific and reasonable evaluation method is particularly important for promoting the sound and steady development of the personal credit market.This paper uses empirical data from personal loan data from Internet credit platforms.Firstly,based on the characteristics of the imbalance of personal loan data,the data is processed in different ways.The personal credit loan model is constructed by the machine learning method based on random forest,and compared with the decision tree model and Logistic model to predict whether the customer has the tendency to default.The work done by the thesis is as follows:Firstly,the theory of machine learning methods and the evaluation criteria for classifiers are mainly introduced in the field of personal credit risk assessment.Secondly,the raw data is preprocessed,including the interpolation of missing values and the processing of outliers.Afterwards,the SMOTE oversampling and random undersampling methods are used to improve the sample imbalance,and the different models are combined with different machine learning algorithms.The AUC values of the individual models were compared.The results show that the SMOTE algorithm deals with the unbalanced credit data and combines with the random forest to perform the best performance,and uses two sets of data sets to verify that the two groups have different degrees of imbalance,and the number of indicators is different,which also better describes the random forest.The practicability and accuracy of the model in personal credit risk assessment and the effectiveness of the SMOTE algorithm in dealing with unbalanced samples.