HDAX: Historical Symbolic Modelling of Delay Time Series in a Communications Network

Homayounfard, H. and Kennedy, P. J.

    There are certain performance parameters like packet delay, delay variation (jitter) and loss, which are decision factors for online quality of service (QoS) traffic routing. Although considerable efforts have been placed on the Internet to assure QoS, the dominant TCP/IP - like the best-effort communications policy - does not provide sufficient guarantee without abrupt change in the protocols. Estimation and forecasting end-to-end delay and its variations are essential tasks in network routing management for detecting anomalies. A large amount of research has been done to provide foreknowledge of network anomalies by characterizing and forecasting delay with numerical forecasting methods. However, the methods are time consuming and not efficient for real-time application when dealing with large online datasets. Application is more difficult when the data is missing or not available during online forecasting. Moreover, the time cost in statistical methods for trivial forecasting accuracy is prohibitive. Consequently, many researchers suggest a transition from computing with numbers to the manipulation of perceptions in the form of fuzzy linguistic variables. The current work addresses the issue of defining a delay approximation model for packet switching in communications networks. In particular, we focus on decision-making for smart routing management, which is based on the knowledge provided by data mining (informed) agents. We propose a historical symbolic delay approximation model (HDAX) for delay forecasting. Preliminary experiments with the model show good accuracy in forecasting the delay time-series as well as a reduction in the time cost of the forecasting method. HDAX compares favourably with the competing Autoregressive Moving Average (ARMA) algorithm in terms of execution time and accuracy.
Cite as: Homayounfard, H. and Kennedy, P. J. (2009). HDAX: Historical Symbolic Modelling of Delay Time Series in a Communications Network. In Proc. Australasian Data Mining Conference (AusDM'09) Melbourne, Australia. CRPIT, 101. Kennedy P. J., Ong K. and Christen P. Eds., ACS. 129-138
pdf (from crpit.com) pdf (local if available) BibTeX EndNote GS