Implementación del marco metodológico CRISP-DM en el análisis y monitoreo de tráfico de red con big data y aprendizaje automático
Mundo FESC · 01.01.2027 · 17 (37)
DOI: 10.61799/2216-0388.2323 ↗Kurz gefasst
Fachartikel in der Zeitschrift Mundo FESC, online erschienen am 01.01.2027. Verfasst von Jorge Enrique Herrera Rubio und 2 weiteren Autor:innen. Die Arbeit nennt 24 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).
Abstract (Originalsprache)
This paper presents a characterization and monitoring study of a three-segment data network from three capture stations (PC1, PC2, PC3) using Big Data. The methodology applied was CRISP-DM: nine statistical flow features were extracted, a tree ensemble classifier was trained for seven services (DNS, FTP, HTTP, HTTPS, ICMP, SMTP, VoIP), and Isolation Forest was applied to 29,720 flows for unsupervised anomaly detection. The results show that the ICMP RTT meets the threshold of ITU-T Recommendation Y.1541 Class 0 (100 ms) at PC1 and PC2, while degradation problems are present at PC3. The classifier achieved 99.59% accuracy, with an F1 macro of 0.988 in holdout and 96.53% accuracy with an F1 macro of 0.913 in the cross-station generalization test. Transactional classes maintained metrics above 0.94. Isolation Forest isolated 591 anomalous flows, reducing inspection complexity. These included a DNS resolution failure and high-volume UDP traffic between specific nodes. Operational findings revealed SYN asymmetry in network segment C and VoIP degradation during routing. It is concluded that the statistical flow results are an active description of the transferable properties of the service itself within the same network, and that CRISP-DM structured the cycle in such a way that the quality defects detected during the scan were transformed into documented preparation rules, laying the foundation for automated network monitoring deployments that disaggregate traffic by service, session, and observation point, thus aiding an analyst in decision-making.
Zitieren
APA
Rubio, J. E. H., Acosta, J. M., & Serrano, B. S. M. (2027). Implementación del marco metodológico CRISP-DM en el análisis y monitoreo de tráfico de red con big data y aprendizaje automático. Mundo FESC, 17(37). https://doi.org/10.61799/2216-0388.2323
Vancouver
Rubio JEH, Acosta JM, Serrano BSM. Implementación del marco metodológico CRISP-DM en el análisis y monitoreo de tráfico de red con big data y aprendizaje automático. Mundo FESC. 2027;17(37). doi:10.61799/2216-0388.2323
BibTeX
@article{Rubio2027,
author = {Rubio, Jorge Enrique Herrera and Acosta, Jairo Moreno and Serrano, Brenda Steffy Moyano},
title = {Implementación del marco metodológico CRISP-DM en el análisis y monitoreo de tráfico de red con big data y aprendizaje automático},
journal = {Mundo FESC},
year = {2027},
volume = {17},
number = {37},
doi = {10.61799/2216-0388.2323}
}
EUVIMED