.png){kind=icon}
.png){kind=icon}
Return
A System for Automatic, Quantitative and Visual Labeling for Failure Management in Cellular Network Data Clusters
.png){kind=icon}
PRE
AI
DOI：10.1109/OJCOMS.2025.3642974.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Cellular network operation strongly relies in the operator's capacity to manage failures and optimize the networks for their efficient and proper functioning. For this, Machine Learning (ML) and Artificial Intelligence (AI) models are deployed to detect and correct problems and inefficiencies in the networks. However, as network operation carries on, new technologies are continuously deployed which, alongside the changes in the networks' environment introduce new unexpected issues and variations in the metrics, reducing the performance of the models. Thus, the used models require being constantly updated, making necessary for operators to optimize their development process. Taking this into consideration, this work proposes a system for labeling clusters with issues based on graphs without prior information. Moreover, as the generated labels are quantitative, they can be used to identify the same issues across several datasets, allowing the application of transfer learning methods to carry knowledge from older datasets to newer ones. The system output has been evaluated using data from two different real-world cellular networks, assessing the capacity of the system to generate accurate and descriptive labels, as well as the labels applicability for transfer learning applications by identifying issues across different datasets.
Keywords:
Labeling
Cellular networks
Measurement
Artificial intelligence
Transfer learning
Principal component analysis
Data models
Training
Dimensionality reduction
Biological system modeling
failure management
transfer learning
labeling
Journal
I
IF:
6.1
Papers:
489
Citations:
0