Implementation of Convolutional Neural Network for Sentiment Analysis on Hotel Customer Reviews

The hotel industry is a business that provides a place to stay for a short period of time. Online reviews provide information about the impressions, evaluations and experiences of hotel customers. The purpose of this study is to classify positive and negative reviews using Convolutional Neural Network method and analyse review sentiments to determine customer perceptions of 6-star hotels in Pekanbaru City based on 8 aspects, namely facilities, rooms, prices, food, services, location, safety and cleanliness. The data source used as a reference is reviews on the Google Maps platform. As a result, the classification using Convolutional Neural Network on customer review data Hotel A for data splitting (60:40), (70:30) and (80:20) has an accuracy of 97%, 98% and 90%. Hotel B has accuracies of 81%, 87% and 85%. Hotel C has an accuracy of 90%, 83% and 88%. Hotel D has an accuracy of 8%, 81% and 89%. Hotel E has an accuracy of 96%, 97% and 98%. Hotel F has an accuracy of 80%, 82% and 81%. The room aspect is the aspect with the most ratings in each hotel rating. The safety aspect is the best aspect in every hotel review, as it is dominated by positive reviews. While the room aspect in every hotel contains many negative reviews and is a material for evaluation, the room aspect in every hotel needs improvement.