Steady Patterns

Abstract : Skypatterns are an elegant answer to the pattern explosion issue, when a set of measures can be provided. Sky-patterns for all possible measure combinations can be explored thanks to recent work on the skypattern cube. However, this leads to too many skypatterns, where it is difficult to quickly identify which ones are more important. First, we introduce a new notion of pattern steadiness which measures the conservation of the skypattern property across the skypattern cube, allowing to see which are the " most universal " skypatterns. Then, we extended this notion to partitions of the dataset, and show in our experiments that this both allows to discover especially stable skypatterns, and identify interesting differences between the partitions.
Complete list of metadatas

Cited literature [9 references]  Display  Hide  Download
Contributor : Alexandre Termier <>
Submitted on : Sunday, December 4, 2016 - 6:30:20 PM
Last modification on : Thursday, February 7, 2019 - 4:16:04 PM
Long-term archiving on : Tuesday, March 21, 2017 - 7:53:33 AM


Files produced by the author(s)


  • HAL Id : hal-01408397, version 1


Willy Ugarte, Alexandre Termier, Miguel Santana. Steady Patterns. Data Science and Big Data Analytics workshop of International Conference on Data Mining, 2016, Barcelone, Spain. ⟨hal-01408397⟩



Record views


Files downloads