Seminar on Computational Learning and Adaptation

  Relational Mining for Temporal Medical Data

Ryutaro Ichise

National Institute of Informatics, Japan

In managing medical data, handling time-series data, which contain irregularities, presents the greatest difficulty. In this talk, I will propose a first-order rule discovery method for handling such data. The method is an attempt to use graph structure to represent time-series data and reduce the graph using specified rules for inducing hypotheses. In order to evaluate the proposed method, I show results from experiments on real-world medical data.

Return to the seminar schedule