Real-Time Popularity Prediction on Instagram

Deming Chu, Zhitao Shen, Yu Zhang, Shiyu Yang, Xuemin Lin

Abstract: Social network services have become a part of modern daily life. Despite explosive growth of social media, people only pay attention to a small fraction of them. Therefore, predicting the popularity of a post in social network becomes an important service and can benefit a series of important applications, such as advertisement delivery, load balancing and personalized recommendation etc. In this demonstration, we develop a real-time popularity prediction system based on user feedback e.g. count of likes. In the proposed system, we develop effective algorithms which utilize the temporal growth of user feedbacks to predict the popularity in real-time manner. Moreover, the system is easy to be adapted for a variety of social network platforms. Using datasets collected from Instagram, we show that the proposed system can perform effective prediction on popularity at early stage of post.