Performance Comparison of Machine Learning-Based Dynamic Resource Allocation Methods for LTE-A Using Realistic Data
No Thumbnail Available
Date
2023-09
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
In the last few years, fourth generation (4G) mobile networks have become
much more complex due to deployments of macro, micro, pico, and femtocells
to boost network capacity and quality of Services(QoS). Long-Term
Evolution Advanced (LTE-A) system is a 4G wireless communication technology
that offers mobile devices high-speed data transmission, increased
capacity, and improved network performance. While LTE-A offers improved
capabilities, it also poses unique resource allocation problems. The
following are some difficulties in allocating LTE-A resources: heterogeneous
network deployment, diverse QoS requirements and dynamic traffic
patterns. Traditional resource management approaches usually use static
strategies or predefined rules and are insensitive to the current demand
on the network. To allocate resources effectively based on the current
network conditions, resource allocation must be adaptable and dynamic.
Adaptive resource allocation methods that can handle the dynamic nature
of LTE-A networks can be developed with the help of machine learning
and artificial intelligence. This thesis evaluates the performance of
machine learning-based resource allocation strategies in LTE-A networks
using realistic data. Four different machine learning-based resource allocation
strategies are compared: Long Short Term Memory (LSTM), Conventional
Neural Network and Long Short Term Memory (CNN-LSTM),
Random Forest (RF) and K Nearest Neighbor(KNN) algorithm. Performance
measures include the accuracy of resource allocation, Root Mean
Square Error (RMSE), Mean Absolute Error(MAE) and speed of convergence(
running time). The results show that LSTM is the model with
higher accuracy score of resource allocation that is 98.56% and in terms
RMSE and MAE random forest has lowest value of 0.294 and 0.08. In the
case of running time KNN takes shortest time of 1.58 second.
Description
Keywords
LTE, LTE-Advanced, Machine Learning, Radio Resource Allocation, Channel quality indicator