Data Analytics for Hotel Reviews in Multi-Language based on Factor Aggregation of Sentiment Polarization

Abstract

Online booking platform eases customer to book hotel easily prior to the arrival date. However, problem arises if customer thinks hotel quality is not as good as promised on online booking platform. Hotel rating which is presented on online booking platform is not sufficient to fully represent hotel quality in terms of services and facilities. To fill in the gap, hotel reviews can be used to depict hotel quality in details. However, hotel reviews can be written in multi-language and the amount is too much, which makes it harder to be understood. On the other hand, hotel’s management also needs to regularly monitor hotel quality perceived by the customer in order to maintain and improve hotel quality. Therefore, this research proposes system to analyze, classify, and predict hotel quality using aggregation of hotel rating and review. The proposed system uses factor aggregation of sentiment polarization approach in vaderSentiment and SentiwordNet which uses different methods (Lexicon-based and Rule-based) to calculate sentiment degree of hotel reviews. Before being analyzed using sentiment analysis, multi-language review is standardized automatically into English. Descriptive and predictive analysis are conducted in big data environment using Hierarchical K-Means Clustering and Smoothing Exponential method respectively. The results of sentiment aggregation that are used as a feature in descriptive and predictive method enriches the analysis for the better result Performance of this approach is 12,47% for prediction error and 4,61% error rate for clustering model. It produced better result hotel segmentation and predictions of hotel quality. This study can give an advantage for both the customer and the hotel management in analyzing the development of hotel quality.