A machine learning pipeline for detecting anomalous energy usage in telecommunications sites

This thesis presents a general framework for identifying telecommunications sites with abnormally high energy consumption. This pipeline is split into three phases. First, data collected from these telecommunications sites is used to train an ensemble of linear regression models that predict energy consumption for a given site. Next, these models are used to generate predictions for sites in the network. These predictions are compared to their ground truth values to generate a set of potential outlier sites. Each of these potential sites is compared against its nearest neighbors to confidently flag the site as an outlier. Finally, anomalous sites alongside useful visualizations are sent to energy management experts so they can manually review the locations and reduce their energy footprint. A baseline instance of this pipeline is implemented to discuss its strengths and limitations.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Computer Science

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2022

Author dc:creator

• Lee, Gregory

Contributors dc:contributor

• Caesar, Matthew

• Machine Learning
• Telecommunications
• Energy

Statement dc:rights

• Copyright 2022 Gregory Lee

Language dc:language

en, eng

Handle dc:identifier

https://hdl.handle.net/2142/117816

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F117816