A distributed Fog node assessment model by using Fuzzy rules learned by XGBoost

The Internet of Things (IoT) will connect more than 50 billion heterogenous devices around the world by 2020. As an Ultra Dense Network (UDN), which needs high resources to be established, different technologies are emerging to improve the efficiency of IoT. Fog is a new phenomenon that uses close powerful nodes to help end users achieve reduced delays, optimize resource consumption, and improve the quality of service. In techniques such as routing, clustering, caching, etc., nodes need to select pairing nodes or the next hop nodes which are used to help nodes transfer or process data. In this paper, a new mathematical fuzzy-based method is proposed to evaluate the suitability of a node’s neighbors. Nodes broadcast their information to inform neighbors about their situations, and each node compares itself to its neighbors, broadcasting a score that shows its tendency to be a pairing node. The proposed method is application-agnostic and can be used in different techniques regarding parameters that are being evaluated. A fuzzy method is used to integrate the parameters and calculate the score. As a new attitude, we use the XGBoost algorithm to extract the fuzzy rules from examples. After receiving the score, another fuzzy method is used to give other eligible neighbors the chance to be the next hop due to support network load balancing. Riverbed Modeler, MATLAB and Python are used to evaluate the node assessment model.