A Packet-Layer Quality Assessment System for VoIP Using Random Forest

In this paper, a novel packet-layer quality assessment system is proposed to monitor the quality of Voice over Internet Protocol services. The efficient machine learning algorithm of random forest is utilized to give the importance of the assessment parameters. The significant parameters are selected to get rid of the disturbance caused by the insignificant ones. To solve the challenge that the usual fitting method is incapable of mapping the complex non-linear correlation between a number of assessment parameters and the quality of voice streaming, the random forest is used again to train the assessment model. The trained model successfully establishes a complex non-linear mapping. The experimental results reveal that the quality assessment model in the proposed system achieves superior performance over the compared models.