Prediction of Link Weight of bitcoin Network by Leveraging the Community Structure

Abstract

In this work, the issue of predicting the edge weight in Bitcoin network has been addressed by leveraging community structure that involves members who trust with each other in their transactions. The proposed model consists of two main stages; the first one is the detection of trusted Bitcoin communities by implementing Newman- Girvan algorithm. In the context, the attributes of node have been modeling in different ways to get different structure of communities each time. Secondly, prediction the missing edge weight based on the neighbors of edge-source in community. In other words, the trust values that pointed to edge-target by neighbors are averaged to represent the prediction of missing edge weight. Practically, the model has been evaluated using two real-world datasets; Bitcoin-OTC and Bitcoin-Alpha datasets. The experimental results explicate the effectiveness of the proposed model comparable with other methods, where the minimization percentage for Bitcoin-OTC dataset is 4% and 18% for all and partial edges respectively. As for Bitcoin-Alpha dataset are 0% and 30% for all and partial edges respectively.