Data mining techniques for prediction and classification in discrete data applications.

摘要

The primary focus of this dissertation is to develop an optimization-based framework for classification and prediction in a variety of application areas. We specifically focused the testing of the framework on automated prediction of stock market trends and on computer-aided medical diagnosis. In order to tune the methods proposed as part of the framework, we have also tested them on other data sets taken from the data mining, AI and machine learning literature. Through extensive computational testing, we show that our framework performs competitively when compared to other prediction methods, in a wide variety of applications.; The main contribution of this dissertation is the advancement of optimization-based techniques for data mining. It is our hope that the success achieved from the combination of these techniques into an integrated framework for classification and prediction will inspire further research at the intersection of data mining and operations research, and will give rise to more applications in industry.