Novel Hybrid Approaches for Occupancy Prediction Using Temperature, Light andCO2 Level Supporting Electrical Energy Management

Authors M. Ennejjar1, M. Ezzini2, M.A. Jallal1,3, S. Chabaa1,4, A. Zeroual1
Affiliations

1I2SP Research Team, Physics Department, Faculty of Sciences Semlalia, Cadi Ayyad University, Morocco

2Fluid Mechanics and Energetic Laboratory, Physics Department, Faculty of Sciences Semlalia, Cadi Ayyad University, Marrakesh, Morocco

3Univ. Grenoble Alpes, CEA, Liten, Campus Ines, 73375, Le Bourget-du-Lac, France

4Industrial Engineering Department, National School of Applied Sciences, Ibn Zohr University, Morocco

Е-mail m.ennejjar.ced@uca.ac.ma
Issue Volume 17, Year 2025, Number 3
Dates Received 23 March 2025; revised manuscript received 24 June 2025; published online 27 June 2025
Citation M. Ennejjar, M. Ezzini, M.A. Jallal, et al., J. Nano- Electron. Phys. 17 No 3, 03038 (2025)
DOI https://doi.org/10.21272/jnep.17(3).03038
PACS Number(s) 07.05.Kf, 88.10.gc
Keywords Occupancy, Energy management, Multi-Layer Perceptron, Hybrid approach, Prediction.
Annotation

This paper introduces novel hybrid approaches for predicting office space occupancy by combining conventional models with artificial neural networks. Specifically, we propose two hybrid models: the Naive Bayes Classifier integrated with a Multi-Layer Perceptron (NBC-MLP) and a Logistic Mixed-Output Perceptron (LMOP). These models use environmental factors such as temperature, light, and CO2 levels to predict occupancy. The hybrid models are designed to leverage the strengths of both conventional models and neural networks, enhancing predictive accuracy while maintaining simplicity. The Naive Bayes Classifier, known for its simplicity with categorical data, complements the Multi-Layer Perceptron’s ability to capture complex relationships in data. The results show that the proposed hybrid models significantly outperform conventional models, with the LMOP model achieving an accuracy of 99.28 %. This indicates the hybrid models’ effectiveness in modeling complex occupancy patterns. Moreover, the models are robust against noisy data and fluctuations in environmental conditions, making them suitable for real-world applications. These models also have practical applications for optimizing space utilization and improving energy efficiency. By predicting occupancy more accurately, they enable better control of HVAC systems and lighting, reducing energy consumption.

List of References