Abstract

Many databases store data in relational format, with differ-ent types of entities and information about links between the entities. The field of statistical-relational learning (SRL) has developed a number of new statistical models for such data. In this paper we focus on learning class-level or first-order dependencies, which model the general database statistics over attributes of linked objects and links (e.g., the percent-age of A grades given in computer science classes). Class-level statistical relationships are important in themselves, and they support applications like policy making, strate-gic planning, and query optimization. Most current SRL methods find class-level dependencies, but their main task is to support instance-level predictions about the attributes or links of specific entities. We focus only on class-level predic-tion, and describe algorithms for learning class-level models that are orders of magnitude faster for this task. Our algo-rithms learn Bayes nets with relational structure, leveraging the efficiency of single-table nonrelational Bayes net learn-ers. An evaluation of our methods on three data sets shows that they are computationally feasible for realistic table sizes, and that the learned structures represent the statistical infor-mation in the databases well. After learning compiles the database statistics into a Bayes net, querying these statistics via Bayes net inference is faster than with SQL queries, and does not depend on the size of the database. 1