Research And Implementation Of Risk Control Model Based On Knowledge Graph

Human activities may always be accompanied by risks,which may violate people's will.Risk has been paid attention to at the beginning of the birth of human civilization.In the second millennium BC,the Code of Hammurabi Code had records on fire insurance and cargo transportation insurance.The risk management field involves millions of aspects.For financial institutions,which is most familiar to people,the risks that may be confronted with mainly include personal credit risk,financial risk in market,and liquidity risk.The credit risk is the most significant.The credit risk mainly refers to the risk that the borrower fails to complete the repayment behavior according to the agreed date of repayment with the lender,resulting in the capital loss of the lender or the corresponding financial institution.This kind of breach of contract has a certain connection with the borrower,whether it is caused by the subjective unwillingness or the objective economic constraints.In this paper,taking P2 P network loan as the background,the research analyzes customers from the data set of loan,builds a complete risk control model,and establishes a risk control platform.The main work includes:(1)Construct the customer's knowledge graph based on the web markup language(schema)developed by foreign search engine companies,the model contains a set of types,organized by multi-level inheritance structure.Using the semi-structured data format JSON-LD to express knowledge map information,organically organize customer information to solve data island problem.(2)The sample data is a type of imbalanced data set.The normal sample is far more than the default sample.It will interfere with algorithm's learning process if all the data are used to be the training data.Traditional under-sampling is a controversial method in dealing with class-imbalance problem because many majority class examples are ignored.To overcome this deficiency,a Clustering-Based Near Miss(CBNM)algorithm was proposed combining the advantages of NearMiss algorithm and K-Means in processing data.CBNM gives a weight to the cluster center by calculating the Near-Miss distance.CBNM algorithm has significant improvement in F-Measure and G-Mean,and the improvement of the classification is also obvious.(3)The traditional risk control model provides a credit reference based on credit rating and default probability.On this basis,a label model derived from rules is established to describe users from multiple aspects and construct customer portraits.Finally,the data after feature processing is compared with the tag data.The experimental results show that the label model can significantly improve the accuracy of the classification task.(4)Complete the prediction of the user's credit rating and whether or not to default.Construct the risk control platform,which can maintain the knowledge graph attribute and search for the customer details according to the label,name,etc.