This paper provides algorithms that use an information-theoretic analysis to learn Bayesian network structures from data. Based on our three-phase learning framework, we develop efficient algorithms that can effectively learn Bayesian networks, requiring only polynomial numbers of conditional independence (CI) tests in typical cases. We provide precise conditions that specify when these algorithms are guaranteed to be correct as well as empirical evidence (from real world applications and simulation tests) that demonstrates that these systems work efficiently and reliably in practice.

This paper addresses the problem of learning Bayesian network structures from data by using an information theoretic dependency analysis approach. Based on our three-phase construction mechanism, two efficient algorithms have been developed. One of our algorithms deals with a special case where the node ordering is given, the algorithm only require ) ( 2 N O CI tests and is correct given that the underlying model is DAG-Faithful [Spirtes et. al., 1996]. The other algorithm deals with the general case and requires ) ( 4 N O conditional independence (CI) tests. It is correct given that the underlying model is monotone DAG-Faithful (see Section 4.4). A system based on these algorithms has been developed and distributed through the Internet. The empirical results show that our approach is efficient and reliable. 1 Introduction The Bayesian network is a powerful knowledge representation and reasoning tool under conditions of uncertainty. A Bayesian network is a directed acyclic graph ...

A framework for integrating methods for decision support; Case-Based Reasoning (CBR) and Data Mining (DM) is outlined. The integration approaches are divided according to which method that is considered to be master and which is the slave. A system using Bayesian networks for computing similarity metrics is implemented and compared to a traditional CBR system. Data are taken from a database from the oil industry. The retrieved cases vary greatly between the systems, especially on features that are unspecified in the "new case". If many features of the "new case" are specified, the new system performs better, according to an evaluation by a domain expert. Introduction Data Mining and Case-Based Reasoning are methods used for decision support ; to organize and process information to make it available for improving the quality of decisions. It is likely that integration of the two methods will lead to a better usage of information. Here, we give a quick introduction to the methods, brief...

A framework for integrating methods for decision support; Case-Based Reasoning (CBR) and Data Mining (DM) is outlined. The integration approaches are divided on which method that is considered to be master and which is the slave. A system using Bayesian networks for computing similarity metrics is implemented and compared to a traditional CBR system. The data that are used are taken from a database from the oil industry. The retrieved cases vary greatly between the systems, especially on features that are unspecified in the "new case". If many features of the "new case" are specified, the new system performs better, according to an evaluation by a database expert. Den som kunne reise, akkurat na... til Paris. Otto Haug, Til Paris Preface In science, one can learn the most by studying what seems the least. Marvin Minsky, The Society of Mind This is a thesis written for obtaining a Masters degree from the department of Computer and Information Science at the Norwegian University o...

In recent years, a number of algorithms have been developed for learning the structure of Bayesian networks from data. In this paper we apply some of these algorithms to a realistic medical domain---stroke. Basically, the domain of stroke is taken as a typical example of a medical domain where much data are available concerning a few hundred patients. Learning the structure of a Bayesian network is known to be hard under these conditions. In this paper, two different structure learning algorithms are compared to each other. A causal model which was constructed with the help of an expert clinician is adopted as the gold standard. The advantages and limitations of various structure-learning algorithms are discussed in the context of the experimental results obtained. Keywords: Bayesian networks, machine learning, knowledge discovery, medical decision support systems. 1