Predictive clustering for structured biomedical data.

In many areas of science, data is generated at an increasing pace, calling for adequate data mining tools to extract useful knowledge. Challenges encountered include the scale and the complexity (structure) of the data. An increasing number of new applications, including the analysis of complex biomedical data, motivate the adaptation of existing, and the design of new methodologies. In this project, we will develop new data mining methods to analyze high-dimensional and structured biomedical data. These methods will be based on predictive clustering trees, a flexible class of machine learning models that can deal with different types of structured data. We will address the tasks of learning predictive bi-clustering trees and learning data representations based on such trees.