Application Of Decision Tree Based Feature Box Algorithm In Credit Scoring Model Of Commercial Banks

With the introduction of China's inclusive financial policy and the rapid development of fintech,the consumer credit field has witnessed explosive growth in recent years.Emerging Internet financial companies have made a breakthrough in the field of credit loans,greatly expanding the breadth and depth of users of financial credit products.At the same time,in the context of increasingly complex credit customer background.Higher requirements are put forward for the risk control and identification ability of financial institutions.Internet finance companies begin to widely use fintech technology represented by machine learning to assist risk control.In this context,commercial Banks have also started a wave of fintech transformation,gradually applying more new technologies and tools to the traditional consumer credit field.In this background,this article through to standard of commercial bank credit rating model on the basis of discussing the development process of against the influence model effect is the most critical step characteristic points method has carried on the exploration,introducing the decision tree algorithm of machine learning in the process of feature points,the gini coefficient for basis points,along with the depth of the tree and the most lobular parameters such as number of nodes,realize the feature points based on decision tree method.In this way,the loss of data information in the feature box is reduced and the prediction effect of the model is improved.The data comparison shows that the decision tree-based partitioning algorithm is significantly improved compared with the traditional equal-frequency partitioning method.In addition,this paper takes a real credit repayment data in the field of Internet finance as the research object,adopts the standard credit score modeling method,and incorporates the decision tree into the feature partitioning algorithm.In the experience of data cleaning,derivative;Key definitions are determined;The credit scoring model was established through sample selection,training set test set classification,feature classification,WOE and IV value calculation,correlation coefficient and multicollinearity test,model establishment,score calibration and other steps,and the effect of the model was evaluated.The results show that the feature partitioning algorithm based on decision tree can fully mine the implicit information of module data and still get quite good prediction effect under the condition of small samples.It provides a novel and effective way for the development of commercial bank credit scoring model.It has good practical application significance.