Article: Predicting the cooling capacity of green buildings using probabilistic neural network models Journal: International Journal of Environment and Pollution (IJEP) 2025 Vol.75 No.4 pp.261 - 279 Abstract: This paper proposes a probabilistic neural network (PNN) model to predict the cooling capacity of green buildings, addressing nonlinear factors and uncertainties often overlooked by traditional regression models. The PNN model uses climate and building features as inputs, applies radial basis function (RBF) in the hidden layer for nonlinear mapping, and generates cooling capacity predictions with confidence intervals. Historical data is used to optimise parameters via backpropagation, and k-fold cross-validation prevents overfitting. Experimental results show that the PNN model achieves an R<SUP align="right"><SMALL>2</SMALL></SUP> value above 0.95 and a 96.67% confidence interval coverage rate across different climate conditions. Compared to traditional models, the PNN demonstrates superior performance in handling nonlinearities and uncertainty in cooling capacity prediction. Inderscience Publishers - linking academia, business and industry through research

Title: Predicting the cooling capacity of green buildings using probabilistic neural network models

Authors: Hua Zheng; Pengming Wang

Addresses: College of Architecture and Energy Engineering, Wenzhou University of Technology, Wenzhou, 325035, Zhejiang, China ' School of Data Science and Artificial Intelligence, Wenzhou University of Technology, Wenzhou 325035, Zhejiang, China

Abstract: This paper proposes a probabilistic neural network (PNN) model to predict the cooling capacity of green buildings, addressing nonlinear factors and uncertainties often overlooked by traditional regression models. The PNN model uses climate and building features as inputs, applies radial basis function (RBF) in the hidden layer for nonlinear mapping, and generates cooling capacity predictions with confidence intervals. Historical data is used to optimise parameters via backpropagation, and k-fold cross-validation prevents overfitting. Experimental results show that the PNN model achieves an R² value above 0.95 and a 96.67% confidence interval coverage rate across different climate conditions. Compared to traditional models, the PNN demonstrates superior performance in handling nonlinearities and uncertainty in cooling capacity prediction.

Keywords: green building cooling capacity prediction; PNN; probabilistic neural network; nonlinear modelling; uncertainty processing; data preprocessing.

DOI: 10.1504/IJEP.2025.150925

International Journal of Environment and Pollution, 2025 Vol.75 No.4, pp.261 - 279

Received: 13 Jan 2025
Accepted: 23 May 2025
Published online: 05 Jan 2026 *

Title: Predicting the cooling capacity of green buildings using probabilistic neural network models

Keep up-to-date