Relational regression: a cognitively-inspired method for prediction system in cognitive IoT

The incorporation of cognition into the design and architecture of the Internet of Things (IoT) has recently given rise to a new subject of IoT research known as cognitive IoT (CIoT). The characteristics and challenges of the IoT are present in the CIoT as well. In CIoT, there are a lot of applications that generate a huge amount of data that require meaningful insight in cognitively-inspired and computationally efficient ways. Therefore, this research introduces relational regression based on the mathematical relation to building a prediction system. The suggested technique begins with extracting the relational pair from large amounts of heterogeneous data and then uses that data for a linear regression model to forecast future sensor readings. The proposed relational regression and prediction system is tested experimentally using 21.25 years of environmental data, with cross-validation performed using several scales. The accuracy measured using mean absolute percentage error and symmetric mean percentage error of the prediction system lies between 99.45% and 99.75%. It shows the efficacy of the proposed relational regression model and prediction system over the competing approaches.