This paper presents an innovative prediction method for key performance indexes with multiple seasonal profiles. The proposed method, called Multiplicative Multi-Seasonal Model (MSMM) relies on a time series decomposition including multiple multiplicative seasonal profiles and a trend component. The method and its underlying model have been specifically designed to be computationally lightweight to scale to big-data scenarios envisaged in upcoming 5G-NFV environments. The MSMM performance has been evaluated on KPI traces of real operating infrastructures/services, made available by Yahoo! The obtained results outlined how the MSMM prediction method provides more accurate forest than well-known algorithm like the seasonal version of ARIMA, with much reduced computational weight.
A lightweight prediction method for scalable analytics of multi-seasonal KPIs
Bruschi R.;Lago P.
2017-01-01
Abstract
This paper presents an innovative prediction method for key performance indexes with multiple seasonal profiles. The proposed method, called Multiplicative Multi-Seasonal Model (MSMM) relies on a time series decomposition including multiple multiplicative seasonal profiles and a trend component. The method and its underlying model have been specifically designed to be computationally lightweight to scale to big-data scenarios envisaged in upcoming 5G-NFV environments. The MSMM performance has been evaluated on KPI traces of real operating infrastructures/services, made available by Yahoo! The obtained results outlined how the MSMM prediction method provides more accurate forest than well-known algorithm like the seasonal version of ARIMA, with much reduced computational weight.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
