Neuro-Ensemble Intelligence for Autonomous Smart Building Ventilation Systems

Abstract

Smart building environments have become increasingly important due to the rising demand for energy efficiency and intelligent infrastructure management. Traditionally, ventilation and environmental control systems have relied on rule-based or threshold-driven mechanisms to regulate indoor conditions such as temperature and CO₂ levels. These conventional heating, ventilation, and airconditioning (HVAC) systems often operate using fixed parameters, which limits their ability to adapt to dynamic environmental changes. As a result, they frequently lead to inefficient energy usage and inconsistent indoor comfort levels. The key challenge lies in the inability of traditional systems to model complex relationships between multiple factors such as temperature, humidity, occupancy, and device activity. Their static control strategies fail to respond effectively to real-time variations, resulting in energy wastage and reduced system performance. This highlights the need for intelligent, data-driven approaches that can improve prediction accuracy and optimize environmental control. To address this issue, the study proposes a smart building environmental analysis system that integrates machine learning models including Support Vector Regression (SVR), K-Nearest Neighbors (KNN), and Gradient Boosting (GB), along with a hybrid RecurrentForest (RCF) model combining Recurrent Neural Network (RNN) and Random Forest (RF). The system includes preprocessing, exploratory analysis, model training, evaluation, and prediction within a unified framework. The proposed RCF model achieved superior performance with R² (Coefficient of Determination) scores of 0.9233 for humidity and 0.9947 for light level. These results demonstrate improved accuracy and reduced error compared to traditional models. The system enhances energy efficiency and indoor comfort, making it suitable for modern smart building applications.