Models from Data for Various Types of Reasoning

Often the primary objective of constructing a model from the data is to model the phenomenonthat produces the data in such a way that the model is useful for the desired type of reasoning objectives. In many graph models of probabilistic knowledge various aspects of these phenomena are represented by nodes while the edges represent the probabilistic dependencies among them. We demonstrate one type of reasoning objective that is better served by those models in which some qualitative relationships of the phenomena, are used as the edges and the hyperedges of the graph models. We have also outlined our methods for handling such models for reasoning and for learning the qualitative relationships of a domain from the empirical data.