DBSCAN-R: A Machine Learning Approach for Routing in Opportunistic Networks

Abstract

Opportunistic networks(OppNets) are a subset of mobile IoT networks in which no direct connection can be established between a message’s source and its destination node. Instead, routing occurs through a collection of intermediate mobile nodes. The lack of a direct connection, along with the mobile nature of the nodes makes routing in OppNets a challenge. This paper aims to utilize machine learning to automate the routing process in Opportunistic networks. The proposed model is a context-aware protocol called DBSCAN-R. DBSCAN-R utilizes DBSCAN(Density-Based Clustering of Application with Noise), an unsupervised soft clustering algorithm to make routing choices. Four dynamic network parameters are chosen to use as features for clustering in DBSCAN. DBSCAN-R outperforms 3 benchmark algorithms i.e Epidemic routing, ProPHET routing, and MaxProp routing when comparing the delivery success rate, average hop count, overhead ratio, and messages dropped.