Selección de Características Para El Modelado de Qos/qoe En Servicios Móviles de Video Y Web: ¿es Beneficiosa La Estructura de Clases Ordinales?

Currently, mobile service providers regard user experience as a fundamental measure of the quality associated with a service. Given a collection of Quality of Service (QoS) indicators, the objective is to estimate the Quality of Experience (QoE), which is assessed through the Mean Opinion Score (MOS). Despite the increasing interest in this subject, several challenges still require further investigation to develop effective strategies that meet users’ expectations regarding service performance. Two major challenges in this domain are: (i) Although QoS/QoE modeling can be approached through an ordinal classification technique, the advantages of this method over conventional classification approaches remain uncertain, and (ii) the relative influence of different QoS parameters in constructing an accurate prediction model is not yet well understood. In this research, we explore the correlation between QoS and QoE in video and web-based services using a machine learning framework. Initially, we contrast the ordinal classification method with the non-ordinal alternative, concluding that the ordinal approach does not inherently provide an advantage, as its effectiveness depends on the specific scenario. Subsequently, we tackle the issue of determining the most significant QoS factors for predicting QoE by implementing feature selection methods. Our findings suggest that not all QoS parameters have the same level of impact on end users’ perception of QoE.