Utilizing Decision Tree-Based Patterns for Predicting Building Energy Consumption

Abstract

Efficiently managing building heating loads (HL) is essential for maximizing energy use. and reducing environmental impact. This study explores the application of decision tree (DT)–based patterns in predicting HL, coupled with two innovative optimizers, the Cheetah Optimization Algorithm (COA) and Smell Agent Optimization (SAO). The research leverages the flexibility and interpretability of DT, a machine learning framework, to framework complex relationships between various building parameters and HL. DTs excel at capturing non-linear relationships, making them suitable for such applications. Incorporating the COA and SAO optimizers introduces an element of intelligence into the framework process. Preliminary outcomes indicate that the combination of DT with COA and SAO optimizers significantly improves the accuracy of HL prediction. This enhancement has promising implications for building management systems, allowing for more precise control of heating systems and energy consumption optimization. Significantly, the hybrid DT+SAO (DTSA) framework delivers reliable outcomes for HL prediction, boasting an impressive correlation coefficient (R2) value of 0.996 as well as a low root mean squared error (RMSE) value of 0.657. This study advances the broader field of energy-efficient building regulation by showcasing the potential of machine learning frameworks and intelligent optimization algorithms for accurately forecasting HL.

Keywords

• heating load
• decision tree
• cheetah optimization algorithm
• smell agent optimization