Assessment of VoIP quality using Bayesian networks

This paper presents a novel method for VoIP quality assessment by means of a Bayesian network. Current quality assessment methods are mainly based on voice stream analysis, their main drawback being the large computing power requirements. This novel method relies on call detail records, fact that substantially reduces the processing power requirements. An analysis of real life data, consisting of 460000 answered calls and 371 customer complaints, was done showing that more than 75% of customer complaints can be identified using this method.