Video Popularity Prediction by Sentiment Propagation via Implicit Network

• DOI: 10.1145/2806416.2806505
• Publication Date: 2016-12-01T14:23:23Z
• Authors: Wanying Ding; Yue Shang; Lifan Guo; Xiaohua Hu; Rui Yan; Tingting He
• Keywords: Popularity; Feature (linguistics); Sentiment Analysis; Univariate

Video popularity prediction plays a foundational role in many aspects of life, such as recommendation systems and investment consulting. Because its technological economic importance, this problem has been extensively studied for years. However, four constraints have limited most related works' usability. First, feature oriented models are inadequate the social media environment, because videos published with no specific content features, strong cast or famous script. Second, studies assume that there is linear correlation existing between view counts from early later days, but not case every scenario. Third, numerous works just take into consideration, discount associated sentiments. Nevertheless, it public opinions directly drive video's final success/failure. Also, approaches rely on network topology, topologies unavailable situations. Here, we propose Dual Sentimental Hawkes Process (DSHP) to cope all problems above. DSHP's innovations reflected three ways: (1) breaks "Linear Correlation" assumption, implements Process; (2) reveals deeper factors affect popularity; (3) topology free. We evaluate DSHP types videos: Movies, TV Episodes, Music Videos, Online News, compare performance against 6 widely used models, including Translation Model, Multiple Linear Regression, KNN ARMA, Reinforced Poisson Process, Univariate Process. Our model outperforms others, which indicates promising application prospect.