Track: Undergraduate Student Paper Competition

Abstract

The application of social media in the field of marketing has grown popular globally. Social media becomes a new preference to many businesses for promoting and advertising. The transformation of conventional marketing into social media is due to its cost-effectiveness. Furthermore, the rapid dispersion of information has led many business people to switch their marketing media into the appropriate instrument by using Twitter. This study implements text mining and k-means for clustering tweets from the Twitter of Indonesian e-commerce, Blibli (@bliblidotcom). This study aims to segment the tweet contents which Blibli can focus on certain contents preferred by Twitter users as their marketing strategies and discover the best formulation of applying k-means. The optimal cluster for k-means accomplished by silhouette method with two distance metrics. The finding of this study provides cosine as the optimal formulation for text clustering problem. The outcome of existing experiments with cosine shows that 15 clusters as the best number. The result of tweet clustering according to the best k-means formulation indicates that the Twitter users tend to like the content about quiz programs named “Fun with Blibli”. Hence, Blibli Indonesia can prioritize that content as marketing strategy in Twitter platform.