A Bayesian Approach to Identify Photos Likely to Be More Popular in Social Media

• 作者: Arunabha Choudhury Sriram Nagaswamy
• 来源期刊: 电脑和通信(英文)
• 学科: 医学
• 关键词: BAYESIAN Supervised Learning Image POPULARITY Classification Data Mining
• 年，卷（期）: 2015,（11）
• 页码范围: 198-204
• 页数: 7页
• 分类号: R73

摘要

With cameras becoming ubiquitous in Smartphones, it has become a very common trend to capture and share moments with friends and family in social media. Arguably, the 2 most relevant factors that contribute to the popularity are: the user’s social aspect and the content of the image (image quality, objects in the image etc.). In recent years, due to various security concerns, it has been increasingly difficult to derive social attributes from social media. Due to this limitation, in this paper we study what make images popular in social media based on the image content alone. We use Bayesian learning approach with variable likelihood function in order to predict image popularity. Our finding shows that a mapping between image content to image popularity can be achieved with a significant recall and precision. We then use our model to predict images that are likely to be more popular from a set of user images which eventually facilitate easy share.