What Users Want for Gig Economy Platforms: Sentiment Analysis Approach

Abstract

Gig economy-based mobile applications are increasingly in demand by the public. An increment in the number of users rises the number of downloads and reviews. However, the number of reviews makes it difficult for developers to understand the information contained in reviews. Besides, one review can have a variety of information. This study proposes a model that can categorize content and sentiment reviews using Support Vector Machine (SVM), Multinomial Naïve Bayes, Complement Naïve Bayes classifier, and Binary Relevance, Classifier Chain, and Label Power Sets as the data transformation method. This study used the reviews contained in the Gojek, Sampingan, and Ruang Guru applications, with 10,123 reviews. This study found the review text’s length influenced accuracy based on the evaluation of Gojek application. Generally, this study results showed that the SVM algorithm (both in the classification of sentiment reviews and review categorization) and Label Power Sets as the transformation method, yielded the best accuracy.