Scholarlink

<p>Occupancy detection is a fundamental capability in intelligent building systems for optimizing energy consumption and enabling responsive environmental control. Conventional machine learning models and standard convolutional neural networks (CNNs) frequently encounter limitations when processing noisy, uncertain, or incomplete environmental sensor data. This work proposes a Neutrosophic Enhanced Convolutional Neural Network (NE-CNN) that integrates neutrosophic preprocessing—decomposing sensor values into truth, indeterminacy, and falsity components—to improve feature quality and mitigate the deleterious effects of uncertainty. The NE-CNN architecture embeds custom NeutrosophicConv1D and NeutrosophicMaxPooling1D layers that learn from the triadic representation while preserving robustness to sensor variability. Empirical evaluation on a public occupancy dataset shows NE-CNN attains 94.5% accuracy while delivering high precision, recall, and F1-score, outperforming conventional ML benchmarks. The framework is scalable and adaptable for real-world deployment and establishes a methodology to integrate neutrosophic reasoning into other deep-learning applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Comparative analysis of neutrosophic enhanced convolutional neural network for occupancy detection in smart buildings

  • 期刊论文

  • 发表日期：

    2025年12月8日

查看全文
  • Ranjeeta Mittal,
  • Suresh Kumar,
  • Urvashi Chugh

摘要

Occupancy detection is a fundamental capability in intelligent building systems for optimizing energy consumption and enabling responsive environmental control. Conventional machine learning models and standard convolutional neural networks (CNNs) frequently encounter limitations when processing noisy, uncertain, or incomplete environmental sensor data. This work proposes a Neutrosophic Enhanced Convolutional Neural Network (NE-CNN) that integrates neutrosophic preprocessing—decomposing sensor values into truth, indeterminacy, and falsity components—to improve feature quality and mitigate the deleterious effects of uncertainty. The NE-CNN architecture embeds custom NeutrosophicConv1D and NeutrosophicMaxPooling1D layers that learn from the triadic representation while preserving robustness to sensor variability. Empirical evaluation on a public occupancy dataset shows NE-CNN attains 94.5% accuracy while delivering high precision, recall, and F1-score, outperforming conventional ML benchmarks. The framework is scalable and adaptable for real-world deployment and establishes a methodology to integrate neutrosophic reasoning into other deep-learning applications.