UdN: A Bio-Inspired Data Network for Significant Pattern Extraction in Cognitive Internet of Things

The cognitive Internet of Things (CIoT) is an emerging field that seeks to embed cognitive capabilities into the design and architecture of the Internet of Things (IoT). While inheriting many features and challenges from traditional IoT systems, CIoT demands computationally efficient and relevant abstractions to handle the vast volumes of data generated by its applications. Addressing this need, this research proposes a bio-inspired approach for extracting meaningful patterns from large, heterogeneous datasets. The methodology begins with probabilistic clustering to organize the diverse data, followed by imputation techniques to address missing values. Subsequently, each cluster's plausibility is computed using probabilistic values, and the cluster with the highest plausibility is selected for further analysis. A bio-inspired unbored data network (UdN) is then constructed for this most plausible cluster. Within this network, a boredom parameter is introduced to measure the network's strength in comparison to the original plausible data structure. The most significant pattern is identified by selecting the mean of the values associated with the unique node in the UdN. The proposed approach demonstrates superior effectiveness—achieving an accuracy exceeding 99.50%—in comparison to existing methods. This is validated through rigorous cross-validation across multiple performance metrics using environmental data collected over a span of 21.25 years.