Scholarlink

<p>Incorporating intelligence into the Internet of Things (IoT) design gave rise to a new field known as cognitive IoT (CIoT). CIoT takes on several features and challenges from IoT. Due to the unreliability of the CIoT sensor network, data is incomplete. As a result, a cognitively inspired method is needed to return the missing information from the CIoT network’s heterogeneous sensor data. Therefore, this study offers a novel technique for restoring the missing data. In the proposed design, a probabilistic strategy is presented to find the joint probability distribution from the marginals caused by the heterogeneous sensors. Subsequently, it estimates the maximum a posteriori for prediction, and the missing value is then approximated by multiplying it by the mean of the incomplete sensory matrix. The suggested technique is tested experimentally on environmental data, using various cross-validation measures, revealing its efficacy (accuracy is greater than 99.42% according to unscaled mean bounded relative absolute error) over competing approaches.</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

An Efficient Approach for Missing Data Recovery in Cognitive IoT Sensor Network

  • 期刊论文

  • 发表日期：

    2025年9月2日

查看全文
  • Vidyapati Jha,
  • Priyanka Tripathi

摘要

Incorporating intelligence into the Internet of Things (IoT) design gave rise to a new field known as cognitive IoT (CIoT). CIoT takes on several features and challenges from IoT. Due to the unreliability of the CIoT sensor network, data is incomplete. As a result, a cognitively inspired method is needed to return the missing information from the CIoT network’s heterogeneous sensor data. Therefore, this study offers a novel technique for restoring the missing data. In the proposed design, a probabilistic strategy is presented to find the joint probability distribution from the marginals caused by the heterogeneous sensors. Subsequently, it estimates the maximum a posteriori for prediction, and the missing value is then approximated by multiplying it by the mean of the incomplete sensory matrix. The suggested technique is tested experimentally on environmental data, using various cross-validation measures, revealing its efficacy (accuracy is greater than 99.42% according to unscaled mean bounded relative absolute error) over competing approaches.