APPLICATION OF NAÏVE BAYES CLASSIFIER METHOD FOR PREDICTING CLAIMS IN AUTOMOBILE INSURANCE

Abstract

Insurance companies often experience difficulties in planning funds. This is caused by the risk of uncertainty in life. Not least, the insurance company went bankrupt because they could not pay their obligations. For this reason, insurance companies need to know the right strategy for setting up reserve funds. One solution that can help insurance companies make decisions and determine strategies is to make claim prediction. In this study, the author will use the Naïve Bayes Classifier method to predict claims in automobile insurance. The Naïve Bayes Classifier itself is a simple probability method where the calculations are based on Bayes’ Theorem. The data used is secondary data from Kaggle.com where this data consists of 10,000 samples with 19 features. The prediction results will be divided into two results, namely 0 and 1 where 0 means "no" and 1 means "yes". "yes" or "no" label will inform us whether the customer will claim or not. The data will go through preprocessing in python so that the format is appropriate. The model will be built without feature selection and with feature selection then being compared to determine the best model. Predicted data result will be compared to the actual data and the accuracy of the best model is 82%. Other evaluation method was applied to evaluate how well the model performed by using ROC – AUC score which has a score of 0.87 and 10-fold cross validation which has an average score of 80%. The result of the prediction will help insurance company with underwriting decision and financial approach or planning.