Artificial Intelligence for KPI Prediction in LTE Networks: A Performance Analysis of GRU vs. SARIMA Models

Abstract

This study addresses the forecasting of Key Performance Indicators (KPIs) in LTE net-works through a comparative analysis of advanced machine learning and statistical models, specifically the Gated Recurrent Unit (GRU) and Seasonal Auto-Regressive Integrated Moving Average (SARIMA) models. Using hourly data from a mobile network operator, the analysis identifies and leverages temporal and statistical patterns, including seasonality and trends, within the KPI dataset to enhance model training