Decentralized multiple hypothesis testing in Cognitive IOT using massive heterogeneous data

An emerging area of study known as Cognitive IoT (CIoT) has emerged as a result of recent efforts to include cognition in the design of the Internet of Things (IoT). Several features and challenges from the IoT are carried over to the CIoT. A lot of applications in CIoT generate massive heterogeneous data, and they require inferential tasks with less computational overhead. Therefore, this study suggests a decentralized computing approach to handle multiple hypothesis testing in less computational time. The first stage is to minimize error at the cluster node by total variance regularization through the alternating direction method of multipliers. Subsequently, the fusion center handles the heterogeneity of data, minimizes model error, extracts the most informative data, and adjusts the p -value for multiple hypotheses testing. The experimental evaluation and cross-validation on six-month traces of the carbon monoxide dataset reveal the efficacy of the proposed algorithm over competing approaches.