Data Aestheticization: A Cognitively-Inspired Method for Knowledge Discovery in Cognitive IoT Sensor Network

New research that incorporates cognition into the IoT architecture has given rise to a new branch of the IoT known as cognitive IoT (CIoT). CIoT inherits several features and challenges of IoT. The applications of CIoT often produce massive amounts of data that are diverse, unpredictable, and dependent on the passage of time. Therefore, a computationally efficient method that draws significant insights from these massive datasets is necessary. With this as our primary focus, we have suggested a novel method that begins by passing the sensor data to a total variation regularizer to mitigate the noisy entries and then by transforming this vast heterogeneous data into its equivalent aesthetic sensed pair data. Following this, the plausibility value for each aesthetic pair is computed, and the pair with the highest plausibility value is taken as the most plausible significant data. The suggested method achieves an accuracy range of 99.14–99.53% on the mean absolute percentage error (MAPE) scale, 99.16–99.57% on the symmetric mean absolute percentage error (sMAPE) metric, and uMbRAE of 99.05–99.65%, which demonstrates its superiority over competing approaches, based on a total of 21.25 years of environmental data.