Abstract
This paper presents an approach to predict Quality of Experience (QoE) in WLAN networks based on information extracted from non-intrusive measurements which only require control and management frame capture. This procedure avoids costly equipment and it is forward compatible. Specifically, three measurements have been taken: the relative time duration of clear-to-send messages, the percentage of blocked beacons, and the time beacons are deferred. Once the information is stored, a statistical model is trained to estimate a QoE Key Performance Indicator (KPI). Specifically, this paper predicts a basic QoE KPI, the time-to-connect that a newcomer user will experience. Three well-known methods of predictive analysis have been evaluated: decision trees, neural networks and support vector machines. Results obtained from real network measurements encourage further investigation to estimate more complex QoE performance indicators.