Algorithm Decision Tree in Analysis Social Media Sentiment to Understand Consumer Views of Brands

Abstract

Online marketplaces in Indonesia are growing rapidly and have become one of the main destinations for internet and social media users. Various marketplace services are available and accessed by the majority of Indonesians. However, with this development, consumer satisfaction with marketplace services varies, from positive, to negative, to neutral. Many consumers express their reactions on social media, including Twitter. In this study, an analysis of opinions was conducted on posts by online business customers in Indonesia on Twitter from various consumers. However, due to the large number of comments, it is difficult to conclude the customer's opinion about online shopping sites that offer the best services. Even trending topics on Twitter only display hot topics that are widely discussed without clear conclusions. To classify general opinion data on Twitter from e-commerce sites, the first step is to process tweet data using Rapidminer tools to recognize the tweet data. Then, the decision tree algorithm is used to categorize opinion data. The results showed that using cross-validation, the decision tree algorithm achieved an accuracy of 70.27 percent while using split validation, it achieved an accuracy of 66.95 percent. In this case, better accuracy was achieved using cross-validation. The results of this study can provide useful information for online businesses in Indonesia to improve the quality of their services and increase customer satisfaction. In addition, this study also provides an overview of the importance of utilizing the decision tree algorithm in categorizing opinion data on social media, especially on Twitter, as a tool for analyzing consumer sentiment towards a service or product.