The Improvement of Indonesian News Curator Classification in Twitter

摘要

Indonesian news curators is Indonesian twitter users that have characteristics like: following the development of news article, sharing the news via retweet or copying URL in tweet, or also giving feedbacks about news article in twitter. This users have important role in helping the process of journalistic especially as news sources in twitter. In generally, there are two kind of news curators in twitter. First news story curator, it was twitter user that made the tweet manually. Second news aggregator, it was twitter user that made tweet automatically using third-party tools. Our previous work about this topic, we developed a framework for classifying this twitter user using seven features such as: profile description, follower, location, URLs, mention, retweet and general tweet. The performance of system produced 86.15% accuracy. In this paper, our purpose is to improve the performance of previous framework with adding five features such as website, verified status, hashtag, number of like and number of retweet. The performance of system using supervised machine learning produced 93.66% accuracy, then we optimized it using Feature Subset Selection method and produced 95.55% accuracy. Lastly, we implemented the framework be a web-based tool that can classified Indonesian twitter user into three classes (news story curator, news aggregator, not both) automatically.