Prediction of MPEG Traffic Data Using a Bilinear Recurrent Neural Network with Adaptive Training

A time-series prediction model using a Bilinear Recurrent Neural Network (BRNN) is proposed in this paper. The BRNN model used in this paper is the Multiresolution architecture with an adaptive training mode. The Multiresolution Bilinear Recurrent Neural Network (MBRNN) is based on the BLRNN that has been proven to have robust abilities in modeling and predicting time series. The proposed MBRNN-based predictor is applied to real-time MPEG video traffic data. The performance of the proposed MBRNN-based predictor is evaluated and compared with the conventional MultiLayer Perceptron Type Neural Network (MLPNN)-based predictor and BRNN-based predictor. When compared with the MLPNN-based predictor and the BRNN-based predictor, the proposed MBRNN-based predictor shows significant improvement in terms of the Normalized Mean Square Error (NMSE) criterion.