In this dataset the evaluation scripts, postprocessed data, and video generation files as described in "Uplink vs. Downlink: Machine Learning-based Quality Prediction for HTTP Adaptive Video Streaming" are available. The evaluation scripts include a random forest, lstm, and neural network based prediction for relevant QoE metrics in video streaming. We tackle the initial delay, video quality and quality changes, video phase prediction and stalling. In the postprocessed data, request information and selected app information for more than 13.000 video runs measured from the native YouTube app are available. Furthermore, we artificially generated 9518 random videos as reference.