Electricity usage recommender system with limited input data
| dc.contributor.author | Al-Dara, Yelyzaveta | en |
| dc.date.accessioned | 2024-06-20T10:42:33Z | |
| dc.date.available | 2024-06-20T10:42:33Z | |
| dc.date.issued | 2024-06-20 | |
| dc.identifier.uri | https://repository.ihu.edu.gr//xmlui/handle/11544/30468 | |
| dc.rights | Default License | |
| dc.subject | Recommender system | en |
| dc.subject | RS | en |
| dc.subject | CF | en |
| dc.subject | Collaborative-filtering | en |
| dc.subject | Appliance disaggregation | en |
| dc.subject | Python | en |
| dc.title | Electricity usage recommender system with limited input data | en |
| heal.abstract | Recommender Systems (RS) in the field of smart grids are gaining popularity. They help the consumers to make corrections to their behavior in order to assist them with achieving their goals and interests. However, the creation of a RS requires the availability of some previously existing data that can describe the interests of the users. In real life problems may arise when creating the RS due to the limited data available. In the literature these problems are often referred to as “cold start” problems. In this work, the problem of creating an electricity usage RS, when limited data are available to be used as an input, is addressed. The input data was limited to the values of the whole-house aggregated electrical power consumption measured with a constant frequency. Data that describes the general characteristics of the building was also utilized. This is namely the type, location and the climate of the area where the building is situated. The usage of a disaggregation algorithm was chosen as a first step included at the beginning of the RS. The suggestion is to utilize one of the training-less disaggregation algorithms based on the graph signal processing. Nevertheless, other training-less disaggregation algorithms could also be embedded at this part of the RS. The second part is the RS itself, which is based on the principle of Collaborative-Filtering (CF) RS using the descriptive data to identify the neighboring buildings. The results of the validation of the RS showed that it can generate recommendations with high accuracy in most of the cases. However, the accuracy drops when each of the neighboring buildings was identified based on the match of different descriptive features with the target building. Thus, some suggestions to improve the accuracy are also introduced. | en |
| heal.academicPublisher | IHU | en |
| heal.academicPublisherID | ihu | en_US |
| heal.access | free | en_US |
| heal.advisorName | Tjortjis, Christos | en |
| heal.committeeMemberName | Tjortjis, Christos | en |
| heal.dateAvailable | 2024 | |
| heal.language | en | en_US |
| heal.license | http://creativecommons.org/licenses/by-nc/4.0 | en_US |
| heal.publicationDate | 2024-06-16 | |
| heal.recordProvider | School of Science and Technology, MSc in Information & Communication Technology Systems | en_US |
| heal.type | masterThesis | en_US |
Αρχεία
Πρωτότυπος φάκελος/πακέτο
1 - 1 από 1
Φόρτωση...
- Ονομα:
- ThesisReport_AL-DARA_Yelyzaveta.pdf
- Μέγεθος:
- 4.33 MB
- Μορφότυπο:
- Adobe Portable Document Format
- Περιγραφή:
Φάκελος/Πακέτο αδειών
1 - 1 από 1
Δεν υπάρχει διαθέσιμη μικρογραφία
- Ονομα:
- license.txt
- Μέγεθος:
- 2.58 KB
- Μορφότυπο:
- Item-specific license agreed upon to submission
- Περιγραφή: