Combining multiple features for image popularity prediction in social media

Popularity prediction, aiming at predicting target items' total interactions with users, is a very significant type of problem and has attracted a lot of attention in recent years. It can benefiit a lot of real applications, such as cold-start recommendation [8] and online advertising [4]. The Social Media Prediction Task-1 (SMP-T1) of the ACM Multimedia 2017 Grand Challenge is designed to predict popularity of photos published by users in social media. In this paper, we introduce the method adopted in this contest detailedly. It is mainly based on carefully designed features and selected regression models. We demonstrate the effectiveness of each feature proposed for this task via univariate and ablation tests by employing different models. Based on those results, we further integrate the verified useful features with the best-performing regression model to obtain final prediction results. We participated this contest with the team name "heihei" and ranked in the second place in the final ranking list.