A Novel Hybrid-Jump-Based Sampling Method for Complex Social Networks

Abstract

With the rapid development of the Internet, social media is affecting and changing people’s lives. The research of network community structure based on a large number of complex network data sets is increasingly popular. Due to the large scale of existing social network data and privacy issues, it is hard to analyze the entire network data directly. Therefore, a reliable and effective network sampling method is very important for the actual estimation of online social networks attributes. Existing network sampling methods like metropolis–hasting random walk (MHRW) can obtain unbiased sample sets from relatively large-scale social networks such as Facebook and describe the key features of the original network. Moreover, MHRW uses a proposed distribution function for sampling control, which can guarantee a well-balanced nature of resulting Markov chain. However, MHRW has the defect of partial graph over entry. In this paper, we proposed a new hybrid jump (HJ) sample by introducing an HJ strategy into MHRW during the sampling progress. First, we use a breadth-first search to obtain a data set without repeated node quickly from a list of jump nodes. Moreover, we adopted uniform sample (UNI) to get the average degree of the original network. Then, a 3-D average degree distribution model is designed to determine the optimal value of the jump parameter in HJ. Finally, we set the condition to execute the HJ strategy in each step of sampling progress. The experimental results demonstrate the performance of HJ is better than the other representation sampling methods both in strong-tie networks and weak-tie networks.