Deterioration detection system using local stationary binomial models for telecommunication services

Kazunori Matsumoto, Kazuo Hashimoto

Research output: Contribution to journalArticle

Abstract

This report concerns a system for monitoring the quality of connections in the gateway switches that interconnect telecommunication networks. Although the time series to be monitored is clearly nonstationary, movements that are almost periodic can be seen if a seven-day interval is set as the period. Therefore, this system introduces a new technique, which is called "partition model optimization," for using the Akaike information criterion (AIC) to determine appropriate groups of local stationary models from a quasi-periodic time series. Also, to perform quality management in real time, a high-speed decision technique called the sequential probability ratio test (SPRT) is introduced. The effectiveness of the partition model optimization technique was systematically investigated in advance by using actual traffic data. Not only was decision parameter maintenance unnecessary, but deterioration could also be detected more accurately than with the conventional system.

Original languageEnglish
Pages (from-to)73-82
Number of pages10
JournalElectronics and Communications in Japan, Part II: Electronics (English translation of Denshi Tsushin Gakkai Ronbunshi)
Volume86
Issue number6
DOIs
Publication statusPublished - 2003 Jun
Externally publishedYes

    Fingerprint

Keywords

  • Akaike information criterion
  • Local stationarity
  • Sequential probability ratio test
  • Traffic deterioration decision

ASJC Scopus subject areas

  • Electrical and Electronic Engineering

Cite this