Study on Diagnosis of Daily Electricity Consumption in Hotel Facilities Based on Amalgamation of Gaussian Process Regression and Teaching Learning Based Optimizer | New Visions in Science and Technology Vol. 5

 The increased usage of energy has negative environmental consequences, such as increased greenhouse gas emissions, global warming, and rapid climate change. This study intends to develop machine learning models for predicting hotel electricity use in an attempt to address this global issue. To anticipate daily electricity usage in hotel facilities, the developed model uses a combination of Gaussian process regression and teaching learning based optimizer (GPR-TLBO). In this context, a teaching learning based optimizer is used to overcome the limitations of manual tuning-based intelligent models by adapting the structure of Gaussian process regression to improve its search skills. This entails fine-tuning the kernel function's type and design parameters. The created model is evaluated by comparing it to eight popular machine learning models. Regression trees (RT), support vector machines (SVM), Elman neural networks (ERNN), generalised regression neural networks (GRNN), back propagation artificial neural networks (BPANN), cascade forward neural networks (CFNN), adaptive neuron fuzzy inference system tuned by particle swarm optimizer (ANFIS-PSO), and adaptive neuron fuzzy inference system tuned by genetic algorithm are among the models used (ANFIS-GA). Mean absolute error (MAE), mean absolute percentage error (MAPE), root-mean squared error (RMSE), relative absolute error (RAE), root relative squared error (RRSE), and geometric reliability index are the six performance assessment measures compared (GRI). Based on their claimed accuracies, the average ranking algorithm is used to create a comprehensive evaluation of the nine machine learning models. Spearman's rank correlation coefficient is used to investigate the levels of dependencies between machine learning models. The developed GPR-TLBO model considerably outperformed the machine learning models, with MAE, MAPE, RMSE, RAE, RRSE, and GRI of 223.98, 8.04 percent, 263.37, 0.57, 0.54, and 1.1, respectively, for MAE, MAPE, RMSE, RAE, RRSE, and GRI. Furthermore, the constructed GPR-TLBO model was shown to have a better overall performance, with an average and standard deviation of rankings of one and zero, respectively. The Spearman's correlation research revealed that the room degree day and daily electricity use have higher levels of interdependence.

Author (S) Details

Eslam Mohammed Abdelkader

Structural Engineering Department, Faculty of Engineering, Cairo University, Giza, Egypt.

Abobakr Al-Sakkaf

Department of Architecture & Environmental Planning, College of Engineering & Petroleum, Hadhramout University, Mukalla, Yemen.

Ghasan Alfalah

Department of Architecture & Building Sciences, King Saud University, Riyadh, Saudi Arabia.

View Book :- https://stm.bookpi.org/NVST-V5/article/view/4103