 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
|
|
Large Graph Data MiningFaculty: Christos FaloutsosStudent: Deepayan ChakrabartiGraphs show up in a wide variety of fields: the Internet topology, the World Wide Web, the network of who-trusts-whom of epinions.com, the graph of who-visits-what-website; the list is endless. Analyzing such graphs can give key insights into their properties, and can help predict how they evolve over time. It also allows us to spot "abnormalities" in a graph, which might be interesting in itself (for example, detecting a denial of service attack on the Internet).Our work focusses on answering exactly these questions. We attempt to find patterns which hold in most real graphs. Based on these patterns, we build models which can generate such patterns and thus, fit the real-world graph well. Such a model should be able to match as wide a variety of graphs as possible, using as only as many parameters as necessary. We try to find procedures for estimating the parameters of these models, and try to analytically study them. |
|