Predictive Analysis of Correlated Time Series with Multiple Seasonality Components

Many Key Performance Indicators (KPIs) in complex systems such as cellular networks are presented to the analytics system as a set of cross-correlated time-series. Many of these time-series have multiple inbuilt seasonalities, and have time-shifted inter-dependencies among themselves that are not evident. We present a case study based on a real-life example of an LTE network to better understand the seasonality and predict future values of the KPIs.
 

Dr Prasad Jogalekar works as a Chief Architect of AI/ML for Ericsson, Santa Clara. He holds a PhD in optimization of cloud architectures, and has worked for various startups and Fortune 500 companies in the domain of networking, storage and applying AI/ML techniques for design and performance optimizations.