Research On P2P Pre-loan Credit Risk Assessment Method Based On Improved Random Forest Algorithm

With the rapid development of Internet Finance,people's financial management concepts and demand for loans have also changed.A new pattern of Internet Finance called P2P(Peer to Peer)lending was born and attracted widespread market attention.Due to the lack of corresponding regulatory agencies and the imperfect personal credit reporting system in the early days in China,P2 P customer default events occur frequently,which has adversely affected investors' rights and the normal operation of the platform.How to establish a scientific and relatively complete credit risk evaluation system has become an urgent problem.This paper studies the relevant data mining classification techniques and methods,and then builds a P2 P pre-loan credit risk evaluation system to provide auxiliary decisions for loan approvers.The main contents of this paper are as follows:(1)The tree set trained by the traditional random forest algorithm consists of decision trees with different classification performance and high similarity.This paper proposes an improved random forest algorithm called TRRF(Trees Reduction Random Forest)based on decision trees reduction.The experiment proves that the TRRF algorithm not only has better classification performance than the traditional random forest algorithm,but also reduces the resource consumption caused by model storage and improves the classification prediction efficiency.(2)Single machine is limited by computing power and memory space when processing massive data.This paper designs and implements parallelization of improved algorithms based on the Spark distributed computing framework.The experiment proves that the parallel algorithm has good parallel performance and scalability.(3)Based on the improved algorithm and the Spark distributed computing framework,this paper studies the theory of P2 P pre-loan credit risk assessment,and then designs and implements the evaluation system based on the improved algorithm.Experiments show that the system has better accuracy and validity of evaluation.