Predição de seguro de automóveis no Brasil utilizando inteligência computacional

Abstract

The use of computational intelligence is increasingly present in various situations in our daily lives, including financially with the use of classification and regression. An efficient classification and regression tool is a Support Vector Machine, designed to obtain good generalizability. This dissertation proposes the forecast of auto insurance in Brazil by using computational intelligence. The proposed method collects data through an application developed for smartphones, ranking the characteristics inserted in terms of importance, selecting the ideal number of characteristics, calculating the expected value of insurance using the SVM regression, and finally calculating the desired quantiles using the bootstrap technique. In the experiments, it was found that the method performs better wiht the adopted metrics, than when the regression used as a benchmark.