Abstract
The enormous increase in the number of Internet of Things data sources (sensors, personal devices and embedded systems) has given rise to a huge amount of unnecessary and redundant data being sent to the cloud. This makes the task of processing and storing this volume of information a very hard one. Therefore, data pre-processing and filtering closer to data sources, such as in fog computing, is necessary. In particular, data reduction in the fog nodes may play a significant role in preventing the dramatic decrease in the Cloud of Things performance, especially in energy consumption, storage space, bandwidth and throughput. However, existing solutions are still lacking, considering they do not achieve the optimal data reduction performance especially in terms of delay and accuracy. In this article, we introduce an approach aiming to eliminate useless and redundant data captured by things basing on some intelligent information extraction techniques. We also evaluate the performance of the proposed solution on a real data set sample to demonstrate that it achieves a good features reduction.