Title: Predicting video engagement using heterogeneous DeepWalk
Authors: Iti Chaturvedi - James Cook University (Australia) [presenting]
Abstract: Video engagement is important in online advertisements where there is no physical interaction with the consumer. This can help identify advertisement frauds where a robot runs fake videos behind the name of well-known brands. Engagement can be directly measured as the number of seconds after which a consumer skips an advertisement. Furthermore, we leverage the fact that videos shown on the same channel have similar viewing behavior. Hence, we use a graph-embedding model called DeepWalk to determine the clusters of videos with high engagement in a particular channel. The learned embedding is able to identify viewing patterns of fraud and popular videos. In order to assess the impact of a video we also consider how the view counts increase or decrease over time. This results in a heterogeneous graph where an edge indicates similar video engagement or history of view counts between two videos. Since it is difficult to find labeled samples for `fraud' video, we leverage a one-class model that can determine `fraud' videos with an outlier or abnormal behaviour. The proposed model outperforms baselines in regression error by over 20\%.