Quantitative OWOM Binary Classification Prediction Based On Machine Learning Approaches

As competition in today's major travel sites intensifies,it has become a strong demand for hotel operators to compete for potential consumers and maintain existing users.Due to this,they need to pay close attention to the generation and consumption of OWOM by prosumers since great OWOM for hotels has been proven to bring ex-cellent reputation to hotel managers,resulting in considerable profits at the same time.Given the importance of OWOM,it would be essential for them to which aspects of reviewer posters' experiences are associated with the positivity of online reviews.Be-cause of this,we focus on textual content of online reviews since the content eventually reflect reviewer posters' experiences.In this study,we try to predict whether a review can be successfully categorized as a positive review based on the textual content of the review from the qualitive aspect of OWOM,using a series of machine learning tech-niques.Specifically,we investigate the various dimensions of the textual content of online reviews like emotional polarity and emotional subjectivity implied in OWOM.By using five machine learning algorithms to construct two-class classifiers,we try to predict whether the textual content would be associated with the quantitative aspects of OWOM such as review ratings and perceived review helpfulness of OWOM.Our prediction results show that polarity implicit in the textual content and the seven di-mensions of the textual content of reviews can be used to predict review ratings.This provides hotel managers with the micro details and profound significance on how to better satisfy their customers.However,our dimensions could not be used to predict perceived review helpfulness,which suggests that in addition to the cumulative effect in time of the review helpfulness,the current dimensions from the textual content of on-line reviews are not well applied to predict perceived review helpfulness.There might be other undiscovered effective elements left which provide profound impact.