Researchon Credit Risk Pridiction Basedon Whale Optimization Algorithm

The financial industry,in today's society is in the era of big data,even more so,in this environment,the bank will develop financial credit business and "machine learning and intelligent optimization algorithm" closely and effective combination,aims to build by big data intelligent,high automatic and accuracy analysis for the main business goals of the intelligent algorithm model,at the same time make it not only conforms to the characteristics of the financial industry can and big data again closely related basic technology of the application.It has become the mainstream of research on how to dig out the characteristics of the customers with the lowest credit risk from the massive credit user data and identify the types of credit customers by constructing the prediction model to reduce the credit risk.Aiming at the credit data with high-dimensional features,this paper proposes to use whale optimization algorithm to find the optimal solution in the feature space,search the feature combination and select the feature subset that contributes the most to the model according to the credit risk classification of the model.Then according to the characteristics of the model portfolio choice subset model of optimal portfolio model and feature subset variables to establish the BP Neural Network(Back Propagation Neural Network)in the evaluation of credit risk prediction model,and in this article use the whale analysis algorithm of BP Neural Network prediction of initial weights and initial threshold prediction accuracy relations has carried on the analysis and optimization,and basic methods to effectively improve the credit risk assessment is applied to forecast the reliability and accuracy of processing model.The main research contents of this paper are summarized as follows:(1)A credit risk feature selection model based on whale algorithm combined with limit learning machine is proposed.In this paper,the credit data set Home Credit was used to preprocess the data such as missing value,outlaw value,coding and normalization,and then the original credit data set after data preprocessing was selected by combining the whale algorithm(WOA)with the limit learning machine.Initialize the feature subset randomly from the original feature set,use the extreme learning machine(ELM)as the classifier,and use the classification error as the fitness function to evaluate the result of feature selection.The whale algorithm is used for continuous iterative optimization to find the optimal feature subset,namely A group of credit risk feature combinations with the highest classification error accuracy.The initial experimental analysis results show that after the original feature subset selection,can ascend to a higher classification error rate,and credit risk feature subset selection based on WOA-ELM initialization model number can be less,the selected feature subsets of credit risk classification error can higher accuracy,less time-consuming,comprehensive analysis of the performance is optimal(2)A credit risk prediction model based on whale algorithm to optimize BP neural network initial weight threshold is established.BP neural network is used to predictcredit risk,And the whale algorithm is introduced to optimize the initial weight threshold of the BP neural network.The results show that the optimization ability of WOA algorithm is better than PSO algorithm and GA algorithm,its mean square error is smaller,and the convergence speed is faster.The WOA model algorithm is used to carry out the initial weight and threshold model based on BP neural network.Optimized,can make the stable value and credit risk model based on BP model prediction accuracy is higher,at the same time can also make it stronger generalization ability.(3)An optimization scheme of hybrid improved whale algorithm is proposed.Firstly,the initial population is improved,and the initial population quality is improved by the population initialization method based on chaotic map.Then,the distance control parameters in the whale algorithm are changed to nonlinear descending to slow down the convergence speed and avoid falling into the local optimal value.Finally,the probability of random search is greater than the probability of spiral search,and the range of global optimization is improved.The research shows that through classification experiments and comparisons on different benchmark test functions,the improved whale classification algorithm has better convergence classification speed and global whale classification optimization ability than the traditional whale classification algorithm before the improvement.Through research and use the improved global whale classification algorithm to design and optimize the credit risk prediction model of the BP neural network,the experimental results show that the improved whale classification is more accurate.This thesis builds a credit feature selection model and credit risk prediction model based on whale algorithm optimization.Experiments show that after the optimization of the whale algorithm,the number of credit risk features decreases and the accuracy of credit risk prediction is higher.