Modeling Hourly Electricity Consumption in Institutional Buildings: An Interpretable Machine Learning Approach
| dc.contributor.author | Doǧanşahin, Kadir | |
| dc.contributor.author | Yüce, Ali | |
| dc.date.accessioned | 2026-06-19T06:31:56Z | |
| dc.date.available | 2026-06-19T06:31:56Z | |
| dc.date.issued | 2025 | |
| dc.department | Malatya Turgut Özal Üniversitesi | |
| dc.description | 9th International Artificial Intelligence and Data Processing Symposium, IDAP 2025 -- 6 September 2025 through 7 September 2025 -- Malatya -- 215321 | |
| dc.description.abstract | In the context of growing global energy demand and environmental concerns, improving energy efficiency in buildings has become a critical objective. This study investigates the determinants of electricity consumption in a university faculty building by analyzing real-world energy usage data in conjunction with environmental and operational variables. Hourly electricity consumption data were collected via a custom-built energy monitoring system, while corresponding temperature and occupancy metrics were obtained from external databases and scheduling information, respectively. To understand the influence of these variables, a series of predictive models were developed and compared. The modeling process began with a multiple linear regression analysis, followed by decision tree-based methods, including CART and Cat-Boost. The linear regression model achieved a test R2 value of 0.599, indicating moderate explanatory power. The CART model improved this to 0.684, while the Cat-Boost model outperformed both, achieving an R2 of 0.800 with the lowest MAE and RMSE values (6.39 kWh and 9.29 kWh, respectively). To better interpret the role of input features, Accumulated Local Effects (ALE) and SHAP value analyses were employed. These interpretability methods confirmed that occupancy and temperature were the most influential predictors, although their effects were nonlinear and context-dependent. The findings highlight the importance of combining machine learning approaches with explainability techniques to enhance understanding of building energy dynamics. The insights gained from this study can inform data-driven energy management strategies in educational buildings and similar facilities. © 2025 IEEE. | |
| dc.description.sponsorship | Malatya Turgut Özal University, (24G10) | |
| dc.identifier.doi | 10.1109/IDAP68205.2025.11222141 | |
| dc.identifier.isbn | 979-833158990-5 | |
| dc.identifier.scopus | 2-s2.0-105025014463 | |
| dc.identifier.scopusquality | N/A | |
| dc.identifier.uri | https://doi.org/10.1109/IDAP68205.2025.11222141 | |
| dc.identifier.uri | https://hdl.handle.net/20.500.12899/4880 | |
| dc.indekslendigikaynak | Scopus | |
| dc.language.iso | en | |
| dc.publisher | Institute of Electrical and Electronics Engineers Inc. | |
| dc.relation.ispartof | 9th International Artificial Intelligence and Data Processing Symposium, IDAP 2025 | |
| dc.relation.publicationcategory | Konferans Öğesi - Uluslararası - Kurum Öğretim Elemanı | |
| dc.rights | info:eu-repo/semantics/closedAccess | |
| dc.snmz | KA_Scopus_20260612 | |
| dc.subject | Cart Decision Tree | |
| dc.subject | Cat-Boost | |
| dc.subject | Energy Management | |
| dc.subject | Linear Regression | |
| dc.title | Modeling Hourly Electricity Consumption in Institutional Buildings: An Interpretable Machine Learning Approach | |
| dc.type | Conference Object |
