Influence analysis for celebrities via public cloud and social platform

Abstract

Recently, the online social networks have emerged as one of the important platforms for social users. Among millions of users, famous person from entertainment circle arouse our interest. They promote social relationship and establish their reputation via these platforms. To analyze the social influence of entertainment stars, we propose and implement a public cloud based framework to crawl celebrities' social messages from Sina Weibo, store the gathered messages, and conduct various analysis to assess the social influence. It consist of three key components: task generation, resource management and task scheduling, and influence analysis. The task generation is responsible of acquiring celebrities' social accounts and issue crawling tasks. We propose a cross-media method to extract social accounts from webpages. The resource management and task scheduling will dynamic adjust the rented resource to minimize the total computing cost while keeping QoS. We propose a dynamic instance provisioning strategy based on the large deviation principle. The influence analysis will undertake various types of analysis, such as fan count, posting frequency, textual analysis, and so on. More than 10,000 celebrities' microblogs have been gathered so far, and some related gainers, such as celebrities and ad agencies can gain the illumination brought by our analysis.