A novel dual prediction scheme for data communication reduction in IoT-based monitoring systems

Internet of things (IoT) based monitoring systems became commonplace. These systems are built upon a large number of devices and sensors. The data collection task of a large number of sensors and devices in an IoT system includes a massive number of data communications. The more the number of devices, the critical is the network bottleneck. In this context, the dual prediction scheme was proposed as a solution for mitigating the large size of communication volumes. The dual prediction scheme consists of a model for predicting future measurements based on historical data. This model is duplicated on both sides, the edge side (i.e., sensor) and the data collection device (i.e., cluster head). The literature includes several works which proposed many dual prediction schemes based on several techniques such as filters and moving average. The literature does not include utilizing the ensemble learning models. This motivates this work to investigate the gradient boosting regression model’s performance compared to the existing solutions. The proposed and state-of-the-art models are evaluated on a realistic dataset. The obtained results show that the proposed model outperforms the existing dual prediction schemes in terms of communication reduction.

© 2022 ICST Institute for Computer Sciences, Social Informatics and Telecommunications Engineering