Multidimensional temporal mining in clinical data

Abstract

Since medical and healthcare data include temporal trends of clinical symptoms, temporal data mining is one of the most important elements to discover knowledge. In this paper, we propose a three-dimensional trajectories mining method to capture the similarities between temporal trajectories of three selected variables. The method was evaluated on two datasets: one is on chronic hepatitis, and the other is on temporal trends of #orders in our university hospital. The results showed that, compared with conventional studies, the method gave more detailed classification of temporal trends.