Research On Internet Financial Risk Control Model Based On Machine Learning Algorithms

With the development of IT Technology,Internet finance industry has risen rapidly.In 2017,the Internet finance industry ranked second in the number of domestic Unicorn enterprises with a value of more than $1 billion,and ranked first in the total value of more than 700 billion RMB.The whole Internet finance industry is very popular.In less than a year,in 2018,there was a collective "thunderstorm" in Internet financial enterprises.In July alone,more than 130 Internet financial enterprises went bankrupt Under the current situation of continuous "thunderstorm",the development of Internet financial industry is difficult,and at the same time,people's understanding of Internet finance returns to rationality.The development of Internet financial industry does not depend on speculation and high profits,but brings convenience to people's lives.The development of Internet Finance in the field of payment benefits from the improvement of people's payment efficiency.It brings convenience to people's life through two-dimensional code payment,chip card payment and other technologies,which is the core value of Internet finance.The loan transaction is the core business of the Internet financial industry.How to let the new technology of Internet optimize the operation logic of traditional finance,control the risk of borrowing and lending in a better way,make it more convenient for users who need funds and have repayment ability to get loans,and make the funds more efficient and reasonable to be used is an urgent matter.Therefore,this paper aims to use machine learning technology to build an effective risk control model,so as to help internet financial enterprises better control lending risk.The main research work is as follows:Firstly,we extract the sample data of Internet financial platform borrowers from multiple dimensions,including consumer data,operator data,user-related data provided by third parties such as Peer Shield,and then further process the data,extracting useful data for building models through feature engineering.Secondly,the classical machine learning algorithms(Logic Regression,Support Vector Machine and Decision Tree)and ensemble learning algorithms(Random Forest and lightGBM)are used to construct the risk control model,and then further improve the risk control model.Thirdly,the logical regression model,linear support vector machine model,random forest model and lightGBM with better performance are selected from the risk control model for Stacking fusion,which further improves the performance of the model.Fourthly,comparing different machine learning algorithms and risk control models constructed in different ways.We can understand the difference of model performance and find the best risk control model.By studying the application of different types of machine learning algorithms in the Internet financial risk control model,this paper provides a method for the construction of the Internet financial risk control model,so as to help the Internet financial enterprises better control the lending risk.At the same time,it also provides ideas for the selection of model algorithm.In addition,this paper further improves the performance of the wind control model through model fusion,which provides guidance and reference for improving the performance of the model.