Selective hypothesis testing in cognitive IoT sensor network

Objects with the ability to learn, reason, and observe their surroundings have recently been the focus of research in the Internet of Things (IoT). Due to this research advancement, a new field called cognitive IoT (CIoT) has emerged. The cognitive IoT combines the IoT with intelligence so that it can mimic the human brain with the help of its intelligent features. The testing of hypotheses is a key part of many CIoT inferential tasks. When dealing with large and diverse datasets, the problem becomes more complicated. Therefore, this study proposes a novel method for selective hypothesis testing that passes the sensory data to a total variation regularizer to reduce the influence of corrupted entries and employs probabilistic clustering to lessen the negative effects of missing entries. In addition, for each generated cluster, the plausibility value is calculated, and the cluster with the highest plausibility value is chosen. This is how large datasets are compressed into more manageable sets. Following this, we address data heterogeneity by designing suitable copula using the minimum Akaike information criterion (AIC). Further, it conducts selective hypothesis testing (SHT) after calculating the p value and adjusting those p values using various methods. Lastly, this study assesses how well the suggested algorithm performs on environmental data spanning 21.25 years. The results show that, even with extremely large and diverse datasets, the suggested method outperforms competing methods in terms of accuracy that lies between 75 and 93%.