Research On Identifying Potential Risks Of Personal Loans Based On K-means Clustering

The balance of non-performing loans and the rate of non-performing loans are the key indicators to measure the operation of banking industry.Personal loans in banking are mainly divided into housing mortgage,production and operation and consumer loans,and the most important factor to measure the quality of loans is the borrower's ability to repay.As the third line of risk management defense,bank internal audit can only identify credit risk after the substantive default of customers' loans,which has serious lag.Based on the traditional personal loan risk management model,there are many drawbacks,such as customer managers' post-loan management in a mere formality,insufficient sampling coverage of risk management department,audit department's risk identification,and poor sampling pertinence.However,as a supervisory department independent of the business department,it has a variety of data collection and processing channels that the business department does not have.Under this premise,it is of great significance to identify the potential credit risk by using the access authority of the internal audit department of the bank to a large number of data,combining the experience of the formation of historical non-performing loans with data mining.Aiming at this problem,this paper establishes a default index model for a single customer through data mining,and gives the evaluation index of the model.Based on the existing stock transaction data and loan attributes of Bank M,the K-means clustering model is used to identify the potential repayment capability of individual loans in stock and the potential risk of overdue loans.The pre-identification of individual loan risk is realized and the retail of Bank M is realized.The effective forecast and control of personal non-performing loan balance and non-performing loan rate show that the model is effective.Firstly,this paper analyzes the current situation of non-performing loans in domestic banking industry,and points out the existing problems of retail personal loans in China's banking industry.On this basis,the main contents and methods of the research and the significance of the research are put forward.Then the classification of bank stock data attributes and the setting of individual default index model are expounded,and the preconditions for large-scale dynamic data processing are put forward.Then this paper analyzes and simplifies the problem of non-performing loans,and gives reasonable assumptions.On the basis of these hypotheses,the inventory customers are classified by clustering model,and the potential risk customers are identified,and the results of the model are analyzed and evaluated.The test results show that the model is effective.