Scholarlink

<p><i>Background</i> The cognitive Internet of Things (CIoT) marks an advanced evolution of IoT, embedding cognitive capabilities that enable systems to learn, reason, and adapt, thereby transforming conventional devices into autonomous, decision-making entities. Building upon the foundational principles of IoT, CIoT is specifically engineered to tackle challenges associated with large-scale, dynamic, and time-critical data. Its effectiveness depends on the ability to extract actionable knowledge from these datasets through intelligent, adaptive, and resource-efficient analytical techniques. <i>Problem statement</i> Traditional approaches struggle to ensure data reliability and uncover meaningful patterns in heterogeneous, long-term environmental datasets. This gap limits effective decision-making in large-scale CIoT environments. <i>Methodology</i> To analyze an environmental dataset spanning 21.25&#xa0;years, we propose a technique based on retroductive reasoning. The method employs probabilistic clustering to categorize observations and address missing data, while the alternating direction method of multipliers (ADMM) is applied for total variance regularization to mitigate corrupted entries. Retroductive values are computed for each cluster using plausibility measures, converted into binary data, and structured into concept lattices. Weighted concepts are then evaluated, and the most significant one is selected using the mean retroductive value. <i>Results</i> The proposed method enhances both data reliability and pattern discovery. Empirical evaluations (accuracy &gt; 99%) demonstrate that it outperforms existing methodologies across multiple performance metrics, including MAPE, sMAPE, and uMbRAE. <i>Conclusion</i> By combining retroductive reasoning, probabilistic clustering, and concept lattice analysis, the study advances CIoT data analysis and supports more effective decision-making in large-scale IoT environments.</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

Retroductive reasoning: a data-driven intelligent method for knowledge discovery using cognitive IoT

  • 期刊论文

  • 发表日期：

    2025年12月30日

查看全文
  • Vidyapati Jha,
  • Priyanka Tripathi

摘要

Background The cognitive Internet of Things (CIoT) marks an advanced evolution of IoT, embedding cognitive capabilities that enable systems to learn, reason, and adapt, thereby transforming conventional devices into autonomous, decision-making entities. Building upon the foundational principles of IoT, CIoT is specifically engineered to tackle challenges associated with large-scale, dynamic, and time-critical data. Its effectiveness depends on the ability to extract actionable knowledge from these datasets through intelligent, adaptive, and resource-efficient analytical techniques. Problem statement Traditional approaches struggle to ensure data reliability and uncover meaningful patterns in heterogeneous, long-term environmental datasets. This gap limits effective decision-making in large-scale CIoT environments. Methodology To analyze an environmental dataset spanning 21.25 years, we propose a technique based on retroductive reasoning. The method employs probabilistic clustering to categorize observations and address missing data, while the alternating direction method of multipliers (ADMM) is applied for total variance regularization to mitigate corrupted entries. Retroductive values are computed for each cluster using plausibility measures, converted into binary data, and structured into concept lattices. Weighted concepts are then evaluated, and the most significant one is selected using the mean retroductive value. Results The proposed method enhances both data reliability and pattern discovery. Empirical evaluations (accuracy > 99%) demonstrate that it outperforms existing methodologies across multiple performance metrics, including MAPE, sMAPE, and uMbRAE. Conclusion By combining retroductive reasoning, probabilistic clustering, and concept lattice analysis, the study advances CIoT data analysis and supports more effective decision-making in large-scale IoT environments.