A Study On The Risk Assessment Model Of Commercial Banks 'individual Credit

With the improvement of people's living standards and the rapid economic and social development,more and more credit consumption and loans take place around the.The rapid development of these businesses has brought a lot of benefits to the operation of the banking market and occupied an increasingly important role.The credit business itself has various uncontrollable factors.These uncontrollable factors can bring risks.This will bring a certain loss to the development of commercial banking business and do not meet the expected return.Therefore,it is possible to establish a scientific and reasonable personal credit risk model to identify the credit risk of loan customers,select no loan or a small amount of loan based on the assessment of the customer's credit risk,and raise the loan interest rate to avoid possible losses.It's a very important thing for a commercial bank.Firstly,according to the credit requirements of the commercial bank credit business,this paper lists the initial indicators that can comprehensively include all aspects of customer information.On the basis of the German credit data set,through several common methods such as correlation analysis,principal component analysis and T test,the index system that affects individual credit risk is obtained.On the basis of this,the clustering method for the initial parameters of Gaussian hybrid model obtained by using kmeans clustering is proposed.First,K new category indicators are obtained,and then the specific index system is obtained by using non-parameters based on the differences between the indicators.Finally,the prediction accuracy of the classification using K nearest neighbor classification algorithm shows that the classification effect of the parameters obtained by Gaussian mixed clustering optimization is better.This paper uses classification accuracy to evaluate the quality of individual credit risk models,and uses K nearest neighbor algorithm to sample the optimization index system of correlation coefficient method,principal component analysis method,T test method,kmeans-gauss mixed model method.The classification accuracy of each index system is obtained.By changing the size of the K value to obtain different classification accuracy,the classification accuracy of the sample data will also change the size of the K value,find the optimal classification accuracy within a reasonable range of K value,and compare the classification accuracy of the initial indicator system.The results show that the classification accuracy of the index system with fewer indicators after optimization is higher than that of the initial classification accuracy,which is 0.21 %,0.13 %,0.97 %,,and 1.27 %.According to the disadvantages of the low classification efficiency of K nearest neighbor algorithm,the improved classification algorithm is used to search the nearest neighbor.The improved classification algorithm has slightly improved the classification accuracy and time consuming has also been shortened.The comparison of the classification accuracy of the PSO-knn algorithm under the indicator system was analyzed,and the classification accuracy of the initial indicator was improved,with 0.54 %,0.13 %,0.8 %and 1.04 % respectively.The results also show that the knn algorithm and PSO-knn algorithm are an effective personal credit evaluation method.