LARM: A Lifetime Aware Regression Model for Predicting YouTube Video Popularity

Online content popularity prediction provides substantial value to a broad range of applications in the end-to-end social media systems, from network resource allocation to targeted advertising. While using historical popularity can predict the near-term popularity with a reasonable accuracy, the bursty nature of online content popularity evolution makes it difficult to capture the correlation between historical data and future data in the long term. Although various existing efforts have been made toward long-term prediction, they need to accumulate a long enough historical data before the prediction and their model assumptions cannot be applied to the complex YouTube networks with inherent unpredictability. In this paper, we aim to achieve fast prediction of long-term video popularity in the complex YouTube networks. We propose LARM, a lifetime aware regression model, representing the first work that leverages content lifetime to compensate the insufficiency of historical data without assumptions of network structure. The proposed LARM is empowered by a lifetime metric that is both predictable via early-accessible features and adaptable to different observation intervals, as well as a set of specialized regression models to handle different classes of videos with different lifetime. We validate LARM on two YouTube data sets with hourly and daily observation intervals. Experimental results indicate that LARM outperforms several non-trivial baselines from the literature by up to 20% and 18% of prediction error reduction in the two data sets.