Abstract:
The identification of "abnormal" events in datasets has long been a focus ofmachine
learning research. Anomaly detection or outlier detection are terms used to describe
this technique. Grubbs provided the most likely first definition in 1969. Outliers are
observations that appear to diverge significantly from the rest ofthe sample. Although
this criteria remains relevant today, the reason for spotting outliers has changed
dramatically. Because pattern recognition algorithms were quite sensitive to outliers
in the data back then, the major reason for the detection was to eliminate the outliers
from the training data thereafter. Data cleaning is another name for this operation. The
interest in anomaly detection has waned significantly with the emergence of more
robust classifiers. However, in the year 2000, academics began to become increasingly
interested in the anomalies themselves, as they are frequently linked to certain
noteworthy occurrences or questionable data sets. Anomaly detection methods are
currently employed in a wide range of applications and are frequently utilised to
supplement classic rule-based detection systems.
Traditionally, the design of a cellular network has focused on energy and resource
optimization to ensure that the network operates smoothly even during peak hours.
This means, however, that radio resources are frequently overprovisioned in cells. In
order to react to changing user demands in the most effective way possible in terms of
energy savings and frequency resource use, next-generation cellular networks require
dynamic management and configuration. Machine Learning approaches are now being
studied in mobile networks to assist with resource management. In this scenario, you
will look at how machine learning can be used to detect abnormal network usage
patterns that might lead to a change in the base station’s setup
