Hybrid Approach to Detect Fault Caused KPIs Anomaly in UMTS Cells
No Thumbnail Available
Date
2020-02-24
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Cellular networks usually suffer from failures or performance degradations due
to several reasons, such as external interference, hardware/software bugs on network
elements, power outages, or cable disconnections. Therefore, to avoid customer
dissatisfaction and loss of revenue due to failures telecom operators need
to detect and respond to performance anomalies of cellular networks instantly.
However, the state of art performance anomaly detection framework (CELLPAD),
which uses a correlation between two Key Performance Indicators (KPI)s as a
means to detect anomaly is not capable of detecting anomalies happing during the
off-pick hour, and; could not differentiate the causes of the anomalies. In this thesis,
we propose a system model, which is capable of detecting anomalies happing
at any traffic load and could differentiate the two causes of a correlational change
anomaly. The proposed system model uses a newly added parameter called mean
Received Total Wideband Power (RTWP) and filtering rules. To assess the performance
of the proposed system model, we conducted an experiment using fourmonth
performance counter data collected from 20
selected sites. The result shows
that the proposed approach improves the detection of sudden drop anomaly by
10
% when compared to the state of the art statistical model, Weighted Moving Average
(WMA). Besides, we can differentiate the two causes of correlational change
anomaly with an F1
-score above 75
%.
Description
Keywords
Anomaly detection, correlational change anomaly, KPI, sudden drop anomaly, RTWP, Universal Mobile Telecommunications System (UMTS)