Learning Bayesian Classifiers from Data

Abstract

Data mining is a new field in computer science that seeks to automatically discover patterns from a large data set for decision making. Machine learning (ML) is a field in artificial intelligence that seeks to improve a computer system's performance by learning from its experience. Both fields apply many theories and techniques developed in statistics. In this talk, I will first present my personal view on the commonalties and the differences between data mining, machine learning and statistics. Next I will introduce learning Bayesian classifiers from Data, an important technique in both ML and data mining. Based on Bayesian statistics, this technique has been shown to outperform competing techniques for applications such as classifying Web sites for search engines. I will discuss its recent development and our initial work on extending this technique to handle continuous variables.