Design and Development of REST-Based Instagram Spam Detector for Indonesian Language

摘要

Instagram, the most popular picture-based social media, has been well known for spam comments, mostly for public figures, including Indonesian actors and actress. Spam comments may lead to misleading information and confusing to other followers. Our previous research on Indonesian language spam comment detector has concluded that K-Nearest Neighbor yields the best result to detect Instagram spam comments. We continue previous our work by designing the REST -based web service to detect Indonesian language spam comments for Instagram. We are using supervised learning combined with Instagram dataset gathered from previous work. This system is built on top of Amazon Web Service (A WS) to provide a robust, resilience, and scalable system using Java Jersey library. The result of this work is a preliminary design and implementation of REST-based Instagram spam detector for Indonesian language that has an average response time of 1678.133 ms with deviation standard of 1178.59 ms. The current service still has limitations in terms of number of features and k-d tree data structure and it still needs some optimization before it can be used in production environment.