Predicting Quality Of Experience For Online Video Systems Using Machine Learning

As the expansion of the online video broadcasting continues in every area of the modern connected world, the need for measuring and predicting the Quality of Experience for content delivery has never been this important. This demo paper has designed and developed a real-time and continuously trained machine learning model in order to predict QoE for online video systems. For this purpose, a platform has been developed where video content is unicasted to a cluster of users simultaneously while objective video metrics are collected into a database. At the end of each video, each user is queried with a subjective survey about their experience. Both quantitative statistics (video metrics) and qualitative information (user surveys) are used continuously as training data to machine learning model. The overall results show that proposed QoE estimation system provides an average Mean Opinion Score (MOS) precision with an error rate ranging from 12% to 15%. This methodology can efficiently answer the problem of predicting user experience for any online video delivery system, while overcoming the problematic interpretation of subjective consumer experience in terms of quantitative metrics.