This repository hosts the code and methodology used in my thesis project, which focuses on the analysis of energy consumption patterns in 5,567 London households. The project aims to explore novel feature extraction techniques and evaluate their effectiveness in identifying distinct consumption patterns.
The analysis is based on an open-source dataset of energy consumption from London households. It can be found here: [https://data.london.gov.uk/dataset/smartmeter-energy-use-data-in-london-households?fbclid=IwAR1H0_wGe8h-sgTUUKwL4B8GUixNBtLcKs8b-PqzupFA6IdHGO2737qLOOI]
We experimented with unique feature extraction methods that are less common in existing literature, aiming to uncover new patterns in energy consumption.
Multiple clustering algorithms were tested, including K-Means++, Fuzzy C-means, etc. Their performance was evaluated based on four clustering validity indexes to ensure robustness in the results.
The final part of the analysis involved applying Ensemble clustering to identify 6 distinct clusters, providing insights into the varying energy consumption patterns among the households.