For Mobile Network Operators to ensure a good experience for their users, it is necessary to have a better idea of what the users actually perceive. The resolution and freezes are objective Quality of Experience(QoE) metrics that give a better indication of the conditions in the video stream rather than using Quality of Service(QoS) values directly.

The purpose of this report was to create a QoE prediction model from LTE system measurements for ultra-low latency live video streams(ULLV). To do this, measurements were performed on a live LTE network using 2 phones. These measurements were analyzed and used in a automated training classification tool to create prediction models. Two models were developed to predict resolutions and freezes. The freezes and resolutions were mapped to a simple QoE indicator(0-10 rating) to mimic a subjective Mean Option Score.

The results show that the predictions models couldn't be used for scheduling purposes in LTE. They can however be used to predict a QoE score of ULLV for monitoring. It achieved a Root Mean Square Error of 1.63 when it was compared to the real resolutions and freezes.