Building time series based prediction / forecast model for a telecommunication network

A system to build a Time Series based prediction/forecast model for a telecommunication network, the system comprising:

• a memory;
• a processor coupled to the memory, wherein the processor is configured to execute programmed instructions stored in the memory to:
• receive historical data, for a predefined period, associated with a telecommunication network, wherein the historical data comprises Performance Management (PM) counters and Key Performance Indicators (KPIs) corresponding to a call-success, a throughput, a processor occupancy, and a time associated with the PM counters and the KPIs;
• pre-process the historical data using statistical and machine learning techniques, wherein the pre-processing of the historical data consists of backfilling missing or anomalous values in the historical data, and evaluating statistical characteristics corresponding to stationarity of the historical data, wherein the pre-processing rectifies unexplained behavior of the historical data that occurred in network level activities, wherein the unexplained behavior of the historical data comprises a step change or anomalies in the KPIs;
• stationarize the historical data by iteratively using techniques of differencing, moving average, and auto-correlation to generate stationarized data;
• assess the KPIs with insight into causation of KPI degradation and tracing the KPIs to a root cause using a Time Series based prediction/forecast model associated with the telecommunication network, wherein the Time Series based prediction/forecast model is built, based on the stationarized data by iteratively determining ‘p’, ‘d’ and ‘q’ parameters using Partial Correlation Function (PACF), differencing, and Auto Correlation Function (ACF) methods, wherein a value of p, a value of d, and a value of q are selected to most accurately model wireless KPIs that exhibit seasonality
• dynamically determine the PM counters and the KPIs for real-time data for a predefined interval using the Time Series based prediction/forecast model;
• instantaneously locate and fix issues related to one or more nodes in the telecommunication network when actual values associated with the PM counters and the KPIs outshoot forecasted values by a predefined threshold value;
• and
• dynamically plan network level activities associated with the telecommunication network in real-time based on the KPIs for the real-time data.

A computer program product having embodied thereon a computer program for building a Time Series based prediction/forecast model for a telecommunication network, the computer program product comprises:

• a program code for receiving historical data, for a predefined period, associated with a telecommunication network, wherein the historical data comprises Performance Management (PM) counters and Key Performance Indicators (KPIs) corresponding to a call-success, a throughput, a processor occupancy, and a time associated with the PM counters and the KPIs;
• a program code for pre-processing the historical data using a machine learning technique, wherein the pre-processing of the historical data consists of backfilling missing or anomalous values in the historical data, and evaluating statistical characteristics corresponding to stationarity of the historical data, wherein the pre-processing rectifies unexplained behavior of the historical data that occurred in network level activities, wherein the unexplained behavior of the historical data comprises a step change or anomalies in the KPIs;
• a program code for stationarizing the historical data by iteratively using techniques of differencing, moving average, and auto-correlation to generate stationarized data;
• and
• a program code for assessing the KPIs with insight into causation of KPI degradation and tracing the KPIs to a root cause using a Time Series based prediction/forecast model associated with the telecommunication network, wherein the Time Series based prediction/forecast model is built, based on the stationarized data by iteratively determining ‘p’, ‘d’ and ‘q’ parameters using Partial Correlation Function (PACF), differencing, and Auto Correlation Function (ACF) methods, wherein a value of p, a value of d, and a value of q are selected to most accurately model wireless KPIs that exhibit seasonality
• a program code for dynamically determining the PM counters and the KPIs for real-time data for a predefined interval using the Time Series based prediction/forecast model;
• a program code for instantaneously locating and fixing issues related to one or more nodes in the telecommunication network when actual values associated with the PM counters and the KPIs outshoot forecasted values by a predefined threshold value;
• and
• a program code for dynamically planning network level activities associated with the telecommunication network in real-time based on the KPIs for the real-time data.