          
          
  
Site maintenance: Ilka
Miloucheva
|
|
Predictive Modelling of Inter-domain QoS – Using ARIMA
with outliers
|
powered
by
Salzburg Research
Uni Budapest |
Goals
- Model validation by policy based monitoring considering different
kind of network connectivity, inter-domain routing policies, costs,
resource and QoS parameter restrictions.
- Long range and short term modelling of inter-domain traffic and
QoS based on monitoring data
- Integration of new algorithms for detecting anomalies in inter-connected
domain infrastructure, normal and abnormal behaviour (“QoS
outliers”) , event pattern study, automated detection and
localisation of specific inter-domain problems (e.g. DoS attacks,
bottleneck and excess capacity)
- Mathematical, operative and predictive models based on measurements
for inter-domain QoS analysis
INTERMON ARIMA based Forecasting with “outliers”
- Modelling and prediction of time series data of QoS parameters
considering “outliers”
- Usage of Box&Jenkins (or ARIMA) methodology (differencing
the process to remove non-stationarity) and preliminary outliers
elimination
- Multivariate QoS outliers – loss and delay
- Comparison of different approaches for forecasting - forecast
without outlier, with correction of outliers and without correction
of outliers with the real data
|