We are pleased to announce that a new version of your favorite social network analysis and visualization software application has been released. SocNetV version 2.2, codenamed "beyond", brings many new features and it is available for Windows, Mac OS X and Linux. Visit the Downloads page to get it!
Here is a brief list of all new features and bugfixes.
Hierarchical Clustering Analysis (HCA)
SocNetV can now perform hierarchical agglomerative cluster analysis on a social network. Supported methods: Single-linkage (minimum), Complete-linkage (maximum) and Average-linkg (UPGMA).
The Structural Equivalence matrix can be computed from the adjacency or the geodesic distances matrix using a user-selected distance metric such as Euclidean distance, Manhattan distance Jaccard distance etc.
The result of the HCA is the list of clusters per level and a dendrogram of the clusters hierarchy in SVG format.
An example screenshot of HCA dendrogram:
Eigenvector Centrality
Version 2.2. adds support for yet another centrality metric: Eigenvector centrality. The Eigenvector centrality of each node in a social network is defined as the ith element of the leading eigenvector of the adjacency matrix. The leading eigenvector is the eigenvector corresponding to the largest positive eigenvalue.
This metric can also be used for embedding radial or level layout in the network.
Pearson product moment correlation coefficients
SocNetV can now correlate actor profiles (ties or distances to other actors) and compute a correlation matrix of pair-wise Pearson coefficients. The computation is done on the given matrix (adjacency or distances) and can be applied on rows (outbound ties), columns (inbound ties) or both.
Actor Similarity
SocNetV can compare the pair-wise tie/distance profiles of actors in order to produce a similarity matrix.
You can select one of the supported measures (Simple Matching, Jaccard, Hamming, Cosine similarity or Euclidean distance).
The algorithm can compare rows (outbound links), columns or both of the input matrix.
Tie profile dissimilarities
SocNetV computes the pair-wise tie profile dissimilarities of the actors, using any of the supported "distance" metrics: Euclidean, Manhattan, Hamming, Jaccard, Chebyshev.
The algorithm uses the adjacency matrix as input. It can compare rows (outbound links), columns (inbound links) or both.
Maximal clique census
SocNetV now uses the Bron–Kerbosch algorithm to find all maximal cliques in an undirected or directed graph.
It produces a census of all MAXIMAL cliques in the network and reports some useful statistics about these.
The clique census report includes disaggregation by vertex and co-membership information. It even prints the cluster analysis dendrogram actor co-membership matrix.
Symmetrize edges by examining Strong Ties
Given a network, the user can create a new relation with only strong ties (when both a->b and b->a exist).
You can opt to create strong ties either from all relations or from current relation only.
Cocitation matrix
SocNetV computes and displays the cocitation matrix of the active social network.
The Cocitation matrix, C=A*A^T, is a NxN matrix where each element (i,j) is the number of actors that have outbound ties/links to both actors i and j.
Cocitation network
When a network is loaded, SocNetV can create a new symmetric relation by connecting actors that are cocitated by others.
In the new relation, an edge will exist between actor i and actor j only if C(i,j) > 0, where C the Cocitation Matrix. Thus the actor pairs cited by more common neighbors will appear with a stronger tie between them than pairs those cited by fewer common neighbors.
Filter (temporarily disable) unilateral (non-reciprocal) edges.
This feature enables you to temporarily symmetrize a directed network by focusing only on the strong, reciprocal ties. Essentially, the program will disable temporarily all unilateral (non-reciprocal) edges.
Multi-relational data read and write in GraphML
SocNetV now supports reading and writing .graphml files with multiple relations. At the moment there is no option as to whether to write one or all relations of the current network to a file. If your network has multiple relations, they will all be saved in the selected GraphML file.
GML format support
The application can load and parse GML formatted network data. At the moment only graph, directed, node, edge, id and labels are supported.
Support for Multirelational directed networks in Pajek files
SocNetV can read Pajek files with multiple relations in *Arcs
Support for EdgeLists with labels
SocNetV can now parse simple and valued EdgeList files where the nodes are referenced by their labels.
HTML reports
SocNetV reports are now saved by default in HTML format. This allows us to have a vastly improved formatting in all reports, as you can see from the screenshots.
Also in Centrality and Prestige reports, you can now sort the results by any column on-the-fly. Just click on the up-down arrow on the header of your preferred column to sort the table by the column contents (ascending or descending).
Note: By default, HTML reports are opened in the system web browser.
Adjacency matrix plot
SocNetV can now plot the adjacency matrix to a file using unicode chars:
Create basic subgraphs with one click
SocNetV can now create clique, star, cycle, and line subgraphs from a group of selected vertices. Just select a group of vertices and select a subgraph from "Selection Subgraph" in the Control Panel or right click on a node and select it in the popup menu (see screenshot below).
Improved performance
SocNetV can multiply matrices faster, by using a recursive Exponentiation by squaring or Fast Modulo multiplication algorithm.
Also, all TextEditor reporting windows will be closed on app exit/init
GUI: Revamped Control Panel and more app icons in the toolbar
The left toolbox (Control Panel) interface was changed to provide a more organized one-click functionality.
The add/remove buttons for nodes and edges have been removed from the Control Panel. They now appear as toolbar icons, along with icons for node finding and properties, edge filtering and application Settings.
The Analyze options in the Control Panel were re-organized in Matrix, Cohesion, Prominence, Communities and Equivalence select popups.
In the Matrix popup list, the user can select which Matrix to compute and view, ie. Adjacency, Cocitation, Laplacian, etc.
The Cohesion popup list includes all basic graph-theory metrics: distances, walks, average distance, clustering coefficient etc.
The Equivalence popup lists all available Structural Equivalence algorithms in SocNetV: HCA, Similarities, Pearson Coefficients, and Tie Profile Dissimilarities.
Other improvements/changes
Select Relation box is now editable. You can change the name of the current relation on the fly.
Also, in case of errors during file importing, the applicaiton informs the user about the line in the file where the errors occurs.
Changes in Datasets
- New dataset: Petersen Graph
- Transformed Krackhardt: High-tech managers and Zachary Karate club into multirelational datasets
Bugs resolved
#1645504 Wrong distances in valued/weighted social networks
#1463087 zero appeal parameter not working in scale free nets
#1629997 Edges with floating point weights are not saved
#1632857 creating nodes randomly may position some of them out of canvas
#1634634 Wrong interpretation of edgedefault graph attribute value
#1636525 Edge labels are not saved in GraphML when the network is undirected
#1637890 Cannot read edgelist files where edges are declared by labels (include non-digit characters)
#1632874 Rubber band selection does not always work
#1633194 Edge arrows option does not exist in Settings
#1633225 Deselect All does not deselect last clicked node
#1622870 Fix large matrix formatting in txt exports
#1637326 Loading multirelational pajek files leaves the app in "not saved" mode
#1 Wrong Group Degree Centralization computation (https://github.com/socnetv/app/issues/1)
#2 Use GitHub as code hosting/repository service (https://github.com/socnetv/app/issues/2)
Important Notices:
Project's new domain: http://socnetv.org
Project's code and files are now hosted in Github. :)
Download and as always: have fun with your social network analysis projects!