Predicting airline customer satisfaction using k-nn ensemble regression models

Abstract

Customer satisfaction questionnaires are a rich and strong source of information for companies to seek loyalty, customer and client retention, opti- mize resources, and repurchase products. Several advanced machine learning and statistical models have been employed to estimate the customer satisfaction score; however, there is not a single model that can yield the best result in all situations. Ensembles of regression techniques have demonstrated their effective- ness for various applications, where the success of these models lies in the con- struction of a set of single models. We perform an experimental study using a real database of 129,890 samples from airline companies, in order to verify the benefits of ensemble models for predicting customer satisfaction. Accordingly, the present paper evaluates the BAGGING ensemble model using the well-renowned k-nn algorithm as the base learner. The obtained results indicate that the BAGGING ensemble performs better than the single classifier in terms of RMSE and MAE.