Aplicação de métodos de mineração de dados em bases de dados de crédito e seguro de clientes

Abstract

For companies, it is important to define the customer categorization system. In this work, two types of customer focus were address, one of which was the automobile insurance market, this service allows a high degree of interaction between company and customer, being consider a high potential market and in an intense growth phase, but the clients can easily switch from insurer depending on your satisfaction. Another focus is that of credit customers, where customers are allowed to borrow from banks depending on their profile, with credit as a means of boosting productive activities. There is a wide range of data from all types of customers, having each branch the need to profile their customers. In order for companies to know what issues are really need for strategic decision-making, the study of Data Mining was apply. The methods used were Random Projection and Principal Component Analysis (PCA), both using the Naive Bayes, J48 and SVM algorithms, with the help of WEKA software. As a result, significant improvements have been shown in the efficiencies of the classifiers involving the methods employed. The Random Projection approach obtained the best results for the two databases analyzed. The J48 and SVM algorithms presented better performance compared to Naive Bayes among the bases. Therefore, from the chosen subsets, they can be submitted to specific analyzes, in order to direct a more precise identification.