Multi-Dimensional Fuzzy Transforms for Attribute Dependencies

We explore attribute dependencies in the datasets by using direct and inverse fuzzy transforms. Our algorithm optimizes the fuzzy partitions of the universe of the attributes and moreover establishes if the set of the data points is sufficiently dense with respect to the chosen partitions: two specific regression indexes measure the reliability of our model. The known “El Nino” dataset is the basis of our experiments, whose results are consistent with the regression analysis made with the same data.