User Financial Portrait Model And Its Feedback Evolution Mechanism Under Big Data

With the development of technologies such as cloud computing,big data,blockchain,and artificial intelligence,financial technology(Fintech)is becoming increasingly hot.Various industries are actively embracing Fintech,and the financial industry is no exception.The traditional financial industry is seeking to cooperate with Internet technology companies and actively embrace financial technology.In the “Internet+” era,the Peer-to-Peer(P2P)network lending market has a high research value as a typical Fintech application scenario.How to assess personal credit risk status comprehensively and accurately is the key to risk control in the financial industry.The traditional financial industry has shortcomings in the comprehensiveness and timeliness of credit assessment.The development of big data analysis and portrait modeling technology provides a new idea for credit evaluation.This article combines big data analysis methods and data portrait modeling techniques.Through the integration and analysis of the borrower's natural attributes and behavioral characteristics on the Renren network lending platform.Consolidate data from various dimensions of borrowers.And a typical model of the user financial portrait under the big data environment——personal credit portrait model is constructed.A large amount of natural attributes and behavioral data left by users borrowing money from the P2 P network on the Internet.This article innovatively proposes to construct user's static portrait model and dynamic portrait model.The static portrait model is a method of mapping the user's Internet behavior into tags.This method is conducive to sort out the user's miscellaneous lending behavior,so that the user's key features are presented succinctly and clearly.The construction of a static portrait model also facilitates the analysis of user static features.The dynamic portrait model is the data fusion of the user's natural attributes and behavioral characteristics.It uses machine learning or deep learning algorithms to conduct evolutionary inference on the borrower's credit.And predicts whether the borrower can repay on time.The static model and the dynamic model act synergistically to analyze the influencing factors of the borrower's borrowing behavior.Constructing the user's static portrait model,this study developed a word segmentation system for natural language processing tasks in the P2 P domain.And use Term Frequency-Inverse Document Frequency(TF-IDF)algorithm to mark the weighted words.Then use the e Xtreme Gradient Boosting(xgboost)algorithm to extract the typical characteristics of the user's behavior.Using these methods can more directly observe the user's behavior characteristics.On the basis of the user's static portrait model,statistical analysis is performed on the dimensions of the user's natural and behavioral characteristics.According to the behavioral economics methodology,the user's group characteristics are verified,which lays a theoretical foundation for the user's dynamic model construction.In order to construct a user's dynamic portrait model,this study uses the xgboost algorithm,bootstrap aggregating(bagging)algorithm,and Convolutional Neural Networks(CNN)to compare the dynamic model construction.According to the research findings,the use of deep learning algorithm has the highest accuracy in the evolutionary reasoning and prediction of user credits based on existing information.Finally,this thesis makes a prospect based on the existing research scenarios and the next research direction.