Learning probabilistic relational models：（学习概率关系模型）.pdfVIP

Learning Probabilistic Relational Models Nir Friedman* Lise Getoor+ Daphne Koller+ Av i Pfeffer+ Hebrew University Stanford University Stanford University Stanford University Abstract example, we may be interested in predicting whether a person is a potential money-launderer based on their bank deposits, A large portio n of real-world data is stored in com- international travel , business connections and arrest records mercial relational database systems. In contrast, of know n associates [Jensen, 1997] . In another case, we may most statistical learning methods wor k only wit h be interested in classifying web pages as belonging to a stu- flat data representations. Thus , to apply these dent, a faculty member, a project , etc. , using attributes of the methods, we are forced to convert our data into web page and of related pages [Craven et al. , 1998] . a flat form , thereby losing much of the relational Unfortunately, few inductive learning algorithms are capa- structure present in our database. This paper builds ble of handling data in its relational form . Most are restricted on the recent wor k on probabilistic relational mod- to dealing wit h a flat set of instances, each wit h its ow n sepa- els (PRMs), and describes how to learn them fro m rate attributes . To use these methods, one typically flattens databases. PRMs allow the properties of an object