Aggregation is one of the Challenges in Learning Probabilistic Relational Models


Abstract:Aggregation is a tool used in presenting the multiple instances of an individual attribute in a single value that characterizes the groups represented. However, probabilistic relational models are constructed from relational data base; these data are interrelated with different cardinalities so it is needed for aggregation in some situations in order to convert the relation cardinality from “many” to “one”. This paper will shed light on the role of aggregation in learning probabilistic relational models through presenting two aggregate functions one is conventional and the other is proposed and compare their effects on the produced models. The results produced from practical work assess that the effect of using different aggregate functions is not determined numerically but conceptually that is needed for intervening the expertise in learning probabilistic relational model.