Customer Risk Assessment And Prediction Based On Financial Big Data

With the rapid development of the Internet finance industry,banks are increasingly using data mining and machine learning technologies.Based on the data such as customer personal information,transaction information and account asset information in the bank database,banks can mine a lot of potential values to improve their business profitability and operational efficiency.Based on the customer data in a commercial bank database,this paper uses the machine learning algorithm to cluster the customers on the basis of the client's portrait.According to the characteristics of different clustering clients,this paper starts from the different attributes of the bank's financial services,demand and value for banks to develop personalized labeling strategy of marketing so as to achieve the purpose of promoting marketing accurately.At the same time,a classification prediction model is constructed to predict the loss of customers in the sample data.According to the bank's actual business needs,this study uses front-end development technology,combined with customer movement trajectory data,from a multi-angle analysis of the bank's big data,and finally designs and implements the bank's big data visualization platform.Specific including:(1)After data preprocessing,machine learning method is adopted to extract characteristic variables that have a great influence on customer behavior analysis,and a method of cluster analysis using K-means clustering algorithm is proposed.Then the clustering results are discriminated and compared by three cases.When the number of clustering k are 5,6 and 7 respectively,the average accuracy are 99.425%,98.717%and 99.671%.Therefore,Under the circumstance that the number of classes is 7 which has the highest accuracy,the corresponding precise marketing strategy is proposed by observing the characteristics of each client group,and finally the visualization of the clustering results is realized.(2)Using the random forest algorithm,Logistic algorithm and decision tree algorithm in machine learning to classify the lost customers and predict the risk clients in the original data,the prediction accuracy and ROC curve of the three algorithms are analyzed.The results show that the prediction accuracy rates of the above three algorithms are respectively 93.2%,88.7%and 90.6%,which shows that random forest algorithm in the bank big data classification prediction is better,and this paper finally predict the results of the visual display.(3)By means of rich visualization tools and customer's moving track,the customer's portrait and banking business are analyzed in all aspects,and the bank data visualization analysis platform was designed and realized.