Project 3Non-overlapping Community Detection in Large NetworksDue: 2019-12-10In this project, you are going to detect / reveal significant communities in largenetworks using current various community detection (graph mining, clustering)methods.1. Detect / reveal communities in the following networks:http://snap.stanford.edu/data/index.htmlNetworks with ground-truth communitiesName Type Nodes Edges Communities Descriptioncom-LiveJournal Undirected,Communities 3,997,962 34,681,189 287,512LiveJournal onlinesocial networkcom-Friendster Undirected,Communities 65,608,366 1,806,067,135 957,154Friendster onlinesocial networkcom-Orkut Undirected,Communities 3,072,441 117,185,083 6,288,363Orkut online socialnetworkcom-Youtube Undirected,Communities 1,134,890 2,987,624 8,385Youtube onlinesocial networkcom-DBLP Undirected,Communities 317,080 1,049,866 13,477DBLP collaborationnetworkcom-Amazon Undirected,Communities 334,863 925,872 75,149 Amazon productnetworkemail-Eu-coreDirected,Communities 1,005 25,571 42 E-mail networkwiki-topcats Directed,Communities 1,791,489 28,511,807 17,364 Wikipedia hyperlinks2. The following methods should be implemented and evaluated:1) Hierarchical clustering using Jaccard index;2) Spectral Clustering;3) CNM (Community Detection in Complex Networks Using External Optimization);4) HRG (Hierarchical random graph, http://tuvalu.santafe.edu/~aaronc/hierarchy/);5) Infomap (http://igraph.org/python/doc/igraph.Graph-class.html);6) Fast Unfolding community detection (http://arxiv.org/pdf/0803.0476v2.pdf,community detection in networkx package);7) Multi-Scale Community Detection using Stability as Optimisation Criterion in aGreedy Algorithm (http://www.elemartelot.org/index.php/programming/cd-code);8) Multi-Scale Community Detection using Stab代写Networks留学生作业、Python编程语言作业调试、Python实验作业代写 代写Database|代做Datility Optimisation(http://www.elemartelot.org/index.php/programming/cd-code);Igraph (http://igraph.org/python) has implemented methods 1)-5);networkx (https://networkx.github.io/,http://blog.sciencenet.cn/blog-404069-337442.html) has implemented 6);Codes of 7) and 8) are listed as above.3. Use “community detection benchmark” to evaluate the performances ofcommunity detection methods:http://arxiv.org/abs/0805.4770https://sites.google.com/site/santofortunato/inthepress2--“Package 1 includes the code to generate undirected and unweighted graphs withoverlapping communities. The extent of the overlap can be tuned by input, and it canbe set to zero if one is interested in non-overlapping clusters.”In this project, we are just focusing on the undirected and unweighted graphs andnon-overlapping communities, such that you can just use the Package1.Note: The Benchmark graphs generates series of graphs with varying degreescommunity structure via changing the mixing parameter μ, for less μ the communitystructure is more distinct and earlier to be detected.4. How to evaluate the performance? The NMI is a good measurement(http://blog.sina.com.cn/s/blog_45e6be080101dlya.html). You can refer to this sourcecode:http://www.mathworks.com/matlabcentral/fileexchange/35625-information-theory-toolbox/content/nmi.m5. Report your experiments’ details, including the methods you used, theimplementation details, and the performance evaluation, comparison, analysis anddiscussion. You are encouraged to improve the current methods, even develop anovel method to apply in these datasets. The report should be consist of four parts:1) introduction and related works; 2) method and implementation; 3) experimentalresults and 4) analysis and discussion.6. You can apply the HPC resource for this project via this link:https://www.must.edu.mo/ssi/labs/hpc.转自：http://www.6daixie.com/contents/3/4733.html
