Critical Algorithm Analysis Based On Cloud Computing Platform For Complex Networks

With the rapid development of the Internet, it brought us into the era of “Knowledge Explosion”, where a huge mount of information makes people convenient at the same time, also increases the difficulty that obtains useful knowledge in massive data. With the arrival of the big data era, complex network, as a new discipline emerged since the 20 th century, went into the path of rapid development, carried out a lot of research in identifying the most influential nodes, community detection, the dynamics theory of transmission in complex network, in which many works have been applied in political, economic, social, physical and biological field. But with the further development, complex network brings us a great challenge that how to analyze the data in large-scale graph. To address these challenge, people began to use the big data tools, such as Hadoop, Spark, GraphLab,for analysis of large-scale complex network research.This paper aims at the study of complex network classic algorithm based on cloud computing. We design and implement three representative complex network algorithms,which is identifying the most influential nodes based on cloud computing, community detection based on cloud computing, SIR propagation based on cloud computing. First of all, after in-depth research in cloud computing platform and classical algorithm of complex network, we put forward some complex network analysis algorithms, which are based on cloud computing platform, for massive graph data. Secondly, according to the characteristics of each algorithm, we choose a appropriate cloud computing platform and design appropriate data structure for algorithms. Different cloud computing platforms have different application scenarios. Finally, many experiments show that these three algorithms are correct in designation and effective in massive graph data.The algorithm of identifying the most influential nodes based on cloud computing is based on leaderrank algorithm, which the node only interact its neighbors, designed in Hadoop and GraphLab platform. The result by testing a small random graph has confirmed this design accuracy. At the same time, the result by testing large-scale graph,such as Twitter, Friendster, has confirmed this design efficiency.The algorithm of community detection based on cloud computing is based on label propagation, designed in GraphLab platform. The result by testing a real graph called Zachary’s Karate Club has confirmed this design valid, and the test by testing large-scale graph, such as com-Orkut, has confirmed this design efficiency.The algorithm of SIR propagation based on cloud computing is based on SIR model and degree infection. Through the experiment, this algorithm has strong flexibility, can be simulated include SI, SIS, SIRS, and shows high scalability in large-scale network.Through a large number of experiments, three algorithms has a great advantage in large-scale graph data.