Name: TULIO VALERIO DA SILVA CORREIA
Publication date: 14/10/2025
Examining board:
Name |
Role |
|---|---|
| FLAVIO MIGUEL VAREJAO | Presidente |
| FRANCISCO DE ASSIS BOLDT | Examinador Externo |
| RODOLFO DA SILVA VILLACA | Examinador Interno |
Summary: In the context of computer networks burdened with heavy applications, large traffic
volumes, and complex topologies, it is crucial to ensure the Quality of Service (QoS) for
user applications, especially with the growing demand for supported services. Recently,
machine learning techniques have been successfully employed to aid in the management
and performance improvement of computer networks. This paper presents an approach that
utilizes data clustering as a method for online sample selection in machine learning-based
monitoring systems, aiming to predict end-to-end service level indicators with low error.
This approach not only optimizes computational resource utilization but also enables
network adaptability to dynamic conditions, while maintaining service quality. The results
demonstrated that the proposed method contributes to the online prediction of key-value
store service read time with competitive error metrics. The normalized mean absolute
error values of the predictions were calculated in the tests conducted throughout the online
sample selection process, yielding an average result of 0.0354, which is a competitive value
compared to the reference methods.
