Cloud-based business intelligence for a cellular IoT network
| Autor: | Johann E.W. Holm, Leon William Moolman, Gabriel Petrus Rossouw van der Merwe |
|---|---|
| Přispěvatelé: | 12868299 - Holm, Johann Erich Wolfgang, 24075477 - Moolman, Liaan W. |
| Jazyk: | angličtina |
| Rok vydání: | 2019 |
| Předmět: |
Service (systems architecture)
Edge device business.industry Computer science 05 social sciences Internet of Things Cloud computing Anomaly detection 050905 science studies computer.software_genre Clustering Health status Business intelligence 0502 economics and business Cellular network Data mining Design science research 0509 other social sciences business Cluster analysis computer 050203 business & management |
| Zdroj: | AFRICON |
| Popis: | This paper presents a cloud-based business intelligence (BI) implementation for a cellular Internet of Things (IoT) network. A Design Science Research (DSR) paradigm, combined with elaborated Action Design Research (eADR) was used to ensure a workable artifact is delivered. The real-world problem is that, in the cellular network considered here, network health status was not initially visible in an intelligent and actionable way. The network health status is used to ensure service availability and includes different health indicators, of which measurements are made at regular intervals. Not all IoT edge devices have health indicators available, but the network under evaluation provided sufficient data from which to extract anomalies. Experiments were conducted to identify the most appropriate anomaly detection technique from three options, namely SARIMA, SVM and LSTM techniques. Anomalies were linked to system operational failures, in turn to be addressed by appropriate standard operating procedures of a larger main-tenance system. Finally, a clustering algorithm was evaluated for automated recognition of anomalous events, showing that anomalies may be clustered in a useful way using the Mean-Shift clustering algorithm, and also identifying additional health indicators that support anomaly classification. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|