Machine Learning Prediction of Consumer Travel Insurance Purchase Behavior

Abstract

Travel insurance covers the expenses and losses related to travelling which is helpful security for travelers domestically or abroad. The goal of the study is to see whether a consumer is interested in buying travel insurance or not based on their information of age, employment type, graduation information, annual income, family size, the existence of chronic diseases, frequent fly or not and ever travel or not which are considered as independent variables. The dataset is collected from an internet website named, Kaggle where some traveler’s information is gathered through a survey of a tourist group. There were 3980 rows of data (Number of Travelers) available, with 9 columns. 10 different types of classification algorithms named logistic regression, KNearest Neighbors (KNN), Gaussian Naive Bayes (Gaussian Naive Bayes), Multinomial Naive Bayes (MNB), Decision Tree Classifier (DT), Random Forest (RF), Support Vector Clustering (SVC), eXtreme Gradient Boosting (XGBoost), Stochastic Gradient Descent (SGD) and Gradient Boosting Classifier (GBC) are performed to select a model with the best accuracy. Among all the algorithms Random Forest, Decision Tree Classifier, and Stochastic Gradient Descent provide the highest accuracy of 88% in predicting whether a consumer would decide to purchase travel insurance or not. The suggested model would be a better choice for insurance companies to make decisions on how to target their desired person and save money with the best profit.