From ScribbleWiki: Analysis of Social Media

Information repositories, link collections

• Collaborative filtering - Data Achieves, Systems and Researchers

Software

Statistical models, including graphical models

• Dynamic Topic Models implementation in C

• Correlated Topic Models implementation in C

• Proximity Software package for relational knowledge discovery (think edge attributes). From UMass.

• MIM Statistical software written for hierarchical models. Similar to R.

Social network computation and analysis

• GMine Developed at CMU, designed for use with large graphs. Works with simple and hierarchical graphs, as well as processing and navigating.

• Netmine Also CMU-based, extracts patterns from networks. Download here. By Deepay Chakrabarti.

• NetworkX- Python add-on package, supports creation of directed and undirected graphs (with simple function calls) and graph properties (degree, diameter, centrality, components, etc). Requires python and a number of other add-ons such as Numeric/numpy, pylab, freetype, and others. From Los Alamos National Labs.

• grTheory- Matlab toolbox, requires optimization toolbox. Not very good for large graphs, as it doesn't include approximations. By Sergiy Iglin.

• JUNG Java Universal Network/Graph Framework. Implements many different algorithms used in social network analysis, also does visualizations. Collaboration of several places.

Visualization

• Graphviz Also known as dotty. Uses very simple language and draws graphs, allows different layouts. Difficult to use for large graphs. By John Ellson and Emden Gansner.

• GUESS Uses Jython (Java + Python) to visualize large networks. Has a nicer interface than graphviz, allows user to change node color, size, and add attributes, etc., and makes pretty pictures. Runs comparatively quickly for larger graphs. (Note, you will also need to know GraphML) By Eytan Adar (UW, formerly HP Labs).

• Prefuse General Java-based data visualization, see the gallery for some examples. Disadvantage is that it is a larger and more difficult to learn package, but it produces better interfaces (for demos and the like). Also it has demos of its own that you can work off of.

• GMine and JUNG See "computation and analysis" section.

Other helpful software

• Tensor Toolbox Ever wondered why Matlab doesn't have sparse multi-dimensional (3D+) array operations? This has them, among other useful tools for tensor analysis. By Tamara Kolda at Sandia National Labs.