学位論文

自己組織型状態空間モデルを用いたネットワークトラヒックの成分分解

jpn

2010-03-24

Not Applicable (or Unknown)

2010-03-24

修士(工学)

工学研究科情報電子工学専攻; 指導教授: 八名和夫

This paper proposes a statistical method for the decomposition of the short term network traffic data. The method introduces a self-organizing state space modeling of the time series. The proposed method has been compared with conventional Kalman filter or the particle filter approach. The new method is capable of describing the characteristic nature of the network traffic data of the mean dependence of fluctuation variance. The proposed method showed better squared error characteristics compared to conventional models. The method will open a new prospect to tie the state space signal modeling with microscopic constructive modeling of the network traffic.

Thesis