Detection of Anomalies in Self-Organizing Networks Traditional versus Modern Machine Learning

PROJECT TITLE :

Anomaly Detection in Self-Organizing Networks Conventional vs. Contemporary Machine Learning

ABSTRACT:

Within the context of anomaly detection in self-organizing networks, this study makes a comparison between traditional Machine Learning and contemporary Machine Learning, also known as Deep Learning. Conventional methods may still offer strong statistical alternatives, particularly when using the appropriate learning representations, despite the fact that Deep Learning has gained a significant amount of traction, particularly in application scenarios where large volumes of data can be collected and processed. Support vector machines, for example, have shown in the past that they have the potential to perform at a state-of-the-art level in a variety of binary classification applications. These machines can also be exploited further by using different representations, such as one-class learning and data augmentation. We show for the first time, using a dataset that has been previously published and is available to the public, that traditional Machine Learning can outperform the previous state-of-the-art method of Deep Learning by an average of 15% across four distinct application scenarios. Our findings also suggest that conventional Machine Learning offers a reliable alternative for 5G self-organizing networks, particularly when the execution and detection times are of the utmost importance. Our findings indicate that conventional Machine Learning offers an improvement in computational speed of nearly two orders of magnitude and a reduction in trainable parameters of one order of magnitude.

Did you like this research project?

To get this research project Guidelines, Training and Code... Click Here