A Comparison of AdaBoost and SVC for Fake Hotel Reviews Detection

Customers increasingly rely on hotel reviews for their bookings whereas hotel administrators use reviews for evaluating their service. Hence, a system that can detect fake reviews can help both customers and hotels administrators obtain accurate information to support their respective decision-making. In this study, we proceed to compare the performance of three machine learning models to classify fake hotel reviews. We used a dataset consisting of a total of 1024 reviews and 1600 pre-labeled reviews from Trip-Advisor. The collected data was preprocessed and split into 80% training data and 20% testing data. The three machine learning models evaluated are AdaBoost classifier (ABC) model with two base estimators: decision tree classifier and Random Forest classifier, and Support Vector classifier (SVC). Training was based on the linguistic features of the training set. During the model validation phase, it is seen that the SVC model provides the best classification performance with an F1-score of 97%. We deduce that SVM can positively contribute to the detection of fake reviews by using the linguistic features of text data.