Name
Abstract | ||
|---|---|---|
Graph construction is a known method of transferring the problem of classic vector data mining to network analysis. The advantage of networks is that the data are extended by links between certain (similar) pairs of data objects, so relationships in the data can then be visualized in a natural way. In this area, there are many algorithms, often with significantly different results. A common problem for all algorithms is to find relationships in data so as to preserve the characteristics related to the internal structure of the data. We present a method of graph construction based on a network reduction algorithm, which is found on analysis of the representativeness of the nodes of the network. It was verified experimentally that this algorithm preserves structural characteristics of the network during the reduction. This approach serves as the basis for our method which does not require any default parameters. In our experiments, we show the comparison of our graph construction method with one well-known method based on the most commonly used approach. |
Year | Venue | Field |
|---|---|---|
2017 | COCOON | k-nearest neighbors algorithm,Data mining,Graph,Computer science,Representativeness heuristic,Network reduction,Network analysis,Data objects,Construction method |
DocType | Citations | PageRank |
Conference | 0 | 0.34 |
References | Authors | |
9 | 3 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Eliška Ochodková | 1 | 30 | 7.54 |
| Sarka Zehnalova | 2 | 8 | 4.83 |
| Milos Kudelka | 3 | 116 | 23.81 |