HIL-Tree: A Hierarchical Structure for Guiding Search into Test and Measurement Data Archives

摘要

This paper describes a novel algorithm that uses discontinuity detection to discover index vectors in test and measurement data archives containing multidimensional data. The index vectors are generated from individual data series in the archive and hold location information about jumps and changes in trends (discontinuities). They are related in a hierarchical manner to form a tree-like structure based on the alignment of the location information across the vectors. We call such trees Hierarchical Index Locations trees (HIL-trees), which are useful in guiding navigation into the raw data and in speeding up the process of retrieving data subsets based on given criteria. To demonstrate the practical value of the algorithm, we present a case study through which the algorithm is applied to real automotive emission test data archives, and show how it works. We also compare the HIL-tree to the well-known R-tree index structure and show how HIL-trees are advantageous in many aspects.