Determining Directions of Service Quality Management using Online Review Mining with Interpretable Machine Learning

Abstract

Determining the importance values of service features is necessary to prioritize the points in service quality management and improvement. Existing studies have used linearly additive relationship models to estimate service feature importance, such as linear and logistic regression. This traditional approach is interpretable but often limited in terms of model fitness and prediction performance. Meanwhile, modern advanced machine learning models provide high fitness and performance but often lack interpretability. Thus, to achieve both reliable prediction and interpretation, we propose a systematic framework for estimating the importance of service features using online review mining with interpretable machine learning. An interpretable machine learning-based method is proposed to estimate the importance values of features by applying the shapley additive global importance metric to the highest-performance prediction model. We validate the superiority of our framework over existing methods through a case study on the global importance estimation of hotel service features in Singapore. To facilitate additional applications, we offer the implementation code of our work at https://github.com/JK-SHIN-PG/OnReviewServImprovement.