Study And Application Of Spark Based Clustering Algorithm For Complex Networks

Nowadays the modern advances and the discovery of new technologies in the science of networks has enlighten our understanding of complex systems.One of the most important features of complex networks is the existence of groups more densely connected than others,or sharing more relations than other groups.These groups are usually called Communities or Clusters.Community detection or clustering in complex networks,can be found in many areas.The role of communities in structuring networks is very important.Thus,traditional methods of classification can be used;especially the construction methods of a partition of the vertices of the graph that maximize some criterion[1].Researchers have proven that it is difficult to categorize the community detection algorithms,but we can categorize them into:hierarchical algorithms,optimization based algorithm(using Newman's modularity),agglomerative algorithms and some use different principles coming from classical clustering like density-based clustering and agent based.In this thesis we introduce community detection in complex network and present some common algorithms related to it.The n we focus our research on two most used algorithms.The first one based on Newman's modularity(Louvain algorithm),and the second based on density clustering(DBSCAN).The fist important goal of this thesis is the implementation of a parallel version of DBSCAN algorithm on top of Apache Spark.Then we proceed to the implementation of a web application for the data analysis and visualization.