Discovering branching and fractional dependencies in databases
- Tsvetanka Georgieva
- Data Knowl. Eng.
This paper explores the use of entropy for visualizing database structure. In particular, we show how visualizing the entropy of a relation provides a global perspective on the distribution of values and helps to identify areas within the relation where interesting relationships may be discovered. The type of structure we are interested in discovering is related to functional dependencies. Our approach is not dependent on the underlying domain of the data, providing a view of the dependency landscape within a relation. Using these techniques we described comparative results for a wide variety of synthetic and real data.