poster
Network Visualisation By Abel B. Nazeh, David T. Jayeoba, Henry P. Ayodele, Luke Collins and Swathi R. Annapureddy Abel B. Nazeh(2038912), David T. Jayeoba (2045578), Henry P. Ayodele(2063341), Luke Collins (1804585), Swathi R. Annapureddy (2046647) L Network Visualisation Tools Presented By: Group 1 Topics of this lecture Complex Networks and their topological properties Tools for visualising a Complex Network Comparisons between Tools Examples of Network Visualisation D What are Complex Networks? Complex Networks are useful tools that visualise data in a connected system They hold a variety of purposes to help establish the key areas effecting a database S Topological properties of a Complex Network Size: The size of a network is measured by its Average Path Length, L: N: Number of Nodes di,j: Geodesic Distance Density: This measures how clustered the data is through the Clustering Coefficient, Ci: ei: Number of Links between neighbours ki: Degree of Node Connectivity: The distribution is based on the probability P(k). The Average Degree however is: M: Number of Edges Tools for visualising Complex Networks NA What is Pajek? It is an open-source program for the analysis and visualization of large networks. It only text files as inputs. Pajek has three versions: Network Source file .CSV Start Convert source file to .Txt using python. Convert .txt file to .Net using Text2Pajek tool. Load .NET data to Pajek interface using provided menu buttons The .Net File Plot Network graph in 2D or 3D Cons: Pajek performs several basic operations on its objects which must be executed in a sequence before results can be obtained. Lack of operating system interoperability, flexibility in input file format and appealing visualizations prevent Pajek from being the top tool for advanced visualizations. Pros: It is the most scalable tool amongst the options for basic visualizations of massive networks with >10 billion nodes. It is memory efficient and very suitable for fast sparse network multiplication. H Gephi support the following file types: GEXF GraphML Pajek NET GDF GML Tulip TLP CSV Compressed ZIP D Gephi is an open-source software used to visualize and manipulate networks. Real-time Visualisation Built-in Rendering Engine Native File Formats Support Layout Algorithm Metrics and Statistics Data Laboratory Dynamic Filtering Data Import and Export Plugins Center Features found in Gephi Easy graph import and export Many options for visual encoding Well defined interfaces Extensible via plugins Open source Pros: Cons: Few visual glitches In-code documentation can be improved What is Gephi? Knime's Pros and Cons Pros: User friendly interface Free community supported software Takes any file type Cons: Node based limitations Difficulty with some topological properties L Knime is an Open-Sourced data analytics, reporting and integration platform. What is Knime? Python is an interpreted, object-oriented, high-level programming language with dynamic semantics. Python software interfaces Python command script, Jupyter notebook, Spyder PyCharm Python Visualisation libraries NetworkX Pyvis Graphviz NetworkKit Topological properties from Python NetworkX graphs and Graphviz calculate the Clustering Coefficient, Average Path Length and Degree of Distribution. Whereas in Pyviz you are unable to calculate these properties. NA Disadvantages of Python: Requires background coding knowledge Python is slower than other coding operators Has an intensive use of memory consumption Advantages of Python: Robust software for large data and professional use Capable of integrating other libraries Many file types are supported Python will allow you to create interactive graphs with animations, effects, etc. What is Python? Comparison between tools Key: * = weaker; ** = medium; *** = good; **** = strongest Features PajekGephiKnimePython Scalability*********** **** User friendliness********* Visual styles *********** Memory efficiency ************ Clustering ******* Manual node/edge editing ********** Layouts******** File formats ************* Speed************** Documentation************* Stability************** Network Profiling* *** *** **** User ratings of the network visualization tools (Pajek, Gephi, Knime and Python) H NA Average Path Length: 2.73 Clustering Coefficient: 0.276 Degree Of Distribution: 6.75 Average Path Length: N/A Clustering Coefficient: 0.701 Degree Of Distribution: 6.579 Average Path Length: 3.52 Clustering Coefficient: 0.275 Degree Of Distribution: 6.64 Average Path Length: 3.90 Clustering Coefficient: 0.899 Degree Of Distribution: N/A Using Game of Thrones Dataset for each Tool Network Visualisation These are network visualization graphs and topological properties of the same dataset but different tools 10 References Knime(2021) KNIME | Open for Innovation Available at: https://www.knime.com/ (accessed by 20, October 2021) W. de Nooy, A. Mrvar, V. Batagelj (2005.) Exploratory Social Network Analysis with Pajek, Structural Analysis in the Social Sciences 27, Cambridge University Press, ISBN:0521602629. Mrvar, A., Batagelj, V. 6 (2016) Analysis and visualization of large networks with program package Pajek: Complex Adapt Syst Model 4, https://doi.org/10.1186/s40294-016-0017-8 (accessed 17, October, 2021) Python (2021) Python documentation available at https://docs.python.org/3/ (accessed by 22, October 2021) Gephi (2021) Gephi software: “Features.” Graph Exploration and Manipulation, available at https://gephi.org/features. D Networkx (2021) Software for Complex Networks — NetworkX 2.6.2 documentation, available at https://networkx.org/documentation/stable/index.html (accessed by 10, October 2021) Thanks for Listening Any questions? By Abel B. Nazeh, David T. Jayeoba, Henry P. Ayodele, Luke Collins and Swathi R. Annapureddy L C:/Documents and Settings/enavarro/Desktop/NetSci2011/poster/poster_netsci2011.dvi Geo-Behavioural Interest Networks GeoBIn Networks Eva M. Navarro-López Alexei Poliakov School of Computer Science The University of Manchester, UK Locomizer www.locomizer.com, UK
[email protected] [email protected] Building a new kind of evolving social networks We are developing a novel framework for the modelling and analysis of peo- ple interaction and interests. It is based on positional information. The challenge is to detect high-correlated groups or pairs of individuals who are not necessarily geographically close. Our starting point is that people’s behaviour can be deter- mined by the types of places they visit or come close to at each time. Inspired by spatial pattern formations in mixtures of cells, we establish links between the position of people and their behaviour From self-organisation of cells to human behaviour Who are you? You are unpredictable You are continually evolving with complex behaviour patterns You have diverse interests and interactions with your environment Surprinsingly! Your behaviour patterns can be defined by simple and universal rules Pattern-formation rules detected in cells of living tissues can explain human behaviour patterns GeoBIn networks: inventing social interest networks Starting point N individuals M places visited Geo-behaviour as an undirected weighted network Modelling the strength and type of geographical and behavioural relationships between individuals IND1 IND2 IND8 IND7 IND6 IND5 IND4 INDNà 1 INDN IND3 ::: :::::: ! 2;6(t) ! 1;2(t) ! 1;N(t) ! 2;5(t) ! 4;5(t) ! 6;7(t) ! 5;8(t) ! 1;7(t) ! 3;7(t) ! 3;N à 1(t) ! 8;N à 1(t) ! 2;4(t) ! 5;N(t) ! 7;N(t) Links define the type of relation- ship Weights for the links are the prox- imity orders of each pair (ωi,j(t)). They are time-varying Proximity orders are obtained from the Geo-Behavioural Interest (GeoBIn) profiles The GeoBIn profile of each individual GeoBIn profile Each person’s GeoBIn profile: is unique contains positional information and proximity indexes is time-varying and evolves with the network Proximity indexes are equivalent to behavioural patterns Evolving characteristics of the GeoBIn network Number of nodes changes with time Number of links changes with time Weights of existing links change with time The network model is behavioural rather than statistical The idea of geo-behaviour was inspired by ... [1 ] W. Taylor, Z. Katsimitsoulia, A. Poliakov. “Simulation of cell movement and interaction”. Journal of Bioinformatics and Computational Biology, 9(1), pp. 91-110, 2011. [2 ] A. Poliakov, M. Cotrina, A. Pasini, D. Wilkinson. “Regulation of EphB2 ac- tivation and cell repulsion by feedback control of the MAPK pathway”. The Journal of Cell Biology, 183(5), pp. 933-47, 2008. This work has been done under the framework of... ‘DYVERSE: A New Kind of Control of Hybrid Systems’ http://www.cs.man.ac.uk/~navarroe/research/dyverse/