Research And Application Of Personal Credit Evaluation Model Based On GBDT And LR Fusion

With the advent of the "Internet Plus" era,the rapid development of China's Internet finance industry has brought about rapid and convenient credit consumption.Credit consumption plays an important role in promoting economic prosperity and development.People's willingness to consume credit is gradually strengthened and their spending power is gradually increasing.Most of the domestic Internet finance industry companies have gradually taken the personal credit consumption business as one of the important areas for follow-up research breakthroughs.However,the scale of personal credit data is gradually expanding.The social data of social platforms and e-commerce data of e-commerce platforms can be used as a part of it.The data type of Internet financial personal credit original data sets is complex and has a large amount of data.The evaluation results of the personal credit evaluation methods of the vast majority of Internet finance industry companies are not very satisfactory,which makes the personal credit consumption business develop slowly.This article addresses the above issues by analyzing the advantages,disadvantages,and complementarities of the GBDT and LR models.The LR linear model has a fast processing speed and good global grasp but requires a relatively high level of features.GBDT is suitable for dealing with non-linear data,and its idea can be used to construct combined features.However,GBDT is not suitable for parallel processing of data sets that are not suitable for processing large amounts of data.A model based on GBDT and LR fusion is proposed.According to the UCI German Credit Data Set,the GBDT model is used to obtain the combined features from the original large amount of data through the experimental single variable method,and the constructed new features are used together with the original data features to use LR for training.By comparing the obtained results with other single models,the credit assessment model based on GBDT and LR fusion has a prediction accuracy rate of87.7%,which is much higher than the single model.The variance of 1.82 indicates that it has certain advantages in terms of stability.This fusion model can be applied widely.Next,the fusion model based on GBDT and LR was applied to the Internet financial dataset.This experiment adopted the "Give me some credit" credit data.First,the Internet finance personal credit evaluation index system was introduced and the data set was performed on this system.Data preprocessing,for the well-processed data sets,establishes a GBDT-based evaluation model,an LR-based evaluation model,a model based on GBDT and LR fusion,and the experimental results show that the AUC value obtained by the fusion model is as high as 0.85.There is a significantincrease in the single model.This paper demonstrates through theory and experiment that the personal credit assessment model based on GBDT and LR has certain advantages in the field of Internet finance personal credit assessment.It is of great practical significance to promote the continuous development and innovation of China's Internet finance industry and is worthy of study.