LTE PRB Utilization Prediction for Load Balancing Between Frequency Layers
No Thumbnail Available
Date
2024-05
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
The ever-increasing number of smart devices and services strains Long-Term Evolution(
LTE) network capacity, impacting Key Performance Indicators(KPIs) like user experience.
Accurate prediction of LTE resource utilization is crucial for network optimization
and improving user experience. Physical Resource Block (PRB) utilization prediction plays
a vital role in analyzing resource allocation within the network. This thesis investigates
the application of machine learning models for predicting LTE PRB utilization to facilitate
load balancing between frequency layers. Three prominent models – Prophet, long
short term memory (LSTM), and eXtreme Gradient Boosting(XGBoost) were evaluated
and compared. The results demonstrate that Prophet significantly outperforms LSTM and
XGBoost in terms of prediction accuracy. Prophet achieved an R-squared value of 0.95 and
a Mean Absolute Error(MAE) of 4.98, indicating a highly accurate fit. Conversely, LSTM
and XGBoost obtained R-squared values of approximately 0.63 with respective MAE values
of around 17. These findings suggest Prophet’s superior accuracy makes it a promising
choice for predicting PRB utilization and enabling effective load balancing in LTE networks.
This thesis contributes to the field of LTE network optimization by demonstrating
the effectiveness of machine learning, particularly Prophet, for PRB utilization prediction.
This capability can be leveraged to develop efficient load balancing algorithms that improve
network performance and user experience.
Description
Keywords
Prophet, LSTM, XGBoost, Physical resource block, MAE, R-squared