Adaptive KNN Classification Algorithm And Its Application In Personal Credit Risk Assessment

The reasonable selection of parameter k is a difficult problem in KNN algorithm design.The k value is generally selected by experience,and all the test samples in the traditional KNN algorithm with a k value,which is obviously unreasonable for samples with uneven distribution and may reduce the accuracy of the classification.Personal credit problem is the main reason for loan defaults,the establishment of an effective personal credit risk assessment system can reduce the bank's investment risk.In order to effectively evaluate the personal credit risk,more and more scholars began to combine data mining technology with personal credit risk assessment to design or develop personal credit risk assessment model based on data mining.But the research on the adaptive mining of personal credit information still lacks the result.This paper mainly studies the adaptive KNN classification algor:ithm and its application in personal credit risk assessment.The main work of this paper is as follows:(1)An adaptive KNN classification algorithm based on local density and purity is proposed to solve the shortcoming of k value in KNN algorithm.The algorithm considers the local density of the test sample and the proportion of the largest class,The test sample selects the k value with a high degree of confidence so that the k value of the test sample is obtained by learning the correlation of the sample,rather than artificially set.For the different test samples selected k value is not fixed,thus improving the accuracy of classification.When requiring a long time experiment to select the k value,use this algorithm can reduce time.(2)In the personal credit risk assessment model,the KNN classification algorithm is introduced.Taking into account the assessment of personal credit,the status of each attribute of the sample is different,there may be some of the characteristics of the credit impact is relatively large,and some characteristics of the impact of credit is small.An improved KNN classification algorithm with feature attribute weighting is proposed,and then the weighted adaptive KNN classification algorithm is applied to the personal credit risk assessment.The experimental results show that the proposed algorithm has a good classification effect on the evaluation of personal credit and can be applied to the assessment of personal credit risk.