Interactive Manipulation, Visualization and Analysis of Large Sets of Multidimensional Time Series in Health Informatics

Abstract

We present a scalable data representation structure which supports interactive access to large sets of time series data, the type of data frequently encountered in health informatics. This structure, called T-Cube, is a cached sufficient statistic equivalent to the data cube concept known in OLAP applications. It differs from a regular data cube in that it can be used to very efficiently answer all conceivable (not just the most common) queries against multidimensional time series databases, including disjunctive queries. Rapid access to complex data not only makes advanced analytics feasible, but it also enables user-level data navigation (drill-downs, roll-ups, visualization) at the interactive speeds. T-Cube also allows for rapid execution of massive screening through data for statistically significant patterns at different levels of aggregation. The exhaustive search strategy guarantees that no event of interest will ever be missed. Successful applications in the domains of public health and food safety indicate that the combined benefits lead to improved situational awareness of the analysts working with information systems powered with T-Cube. We are on the outlook for other applications where it may be of help.