Flexible selection of wavelet coefficients based on the estimation error of predefined queries

  • Authors:

  • Jaehoon Kim;Seog Park

  • Affiliations:

  • Department of Computer Science, Sogang University, Seoul, Korea;Department of Computer Science, Sogang University, Seoul, Korea

  • Venue:
  • PAKDD'07 Proceedings of the 2007 international conference on Emerging technologies in knowledge discovery and data mining
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper, we introduce a data stream reduction method using lossy wavelets compression. The lossy compression means that compressed data carry as much information about the original data stream as possible while the original data size remarkably reduced. We think that wavelets technique should be an efficient method for such lossy compression. Especially we consider storing a plenty of past data stream into stable storage (flash memory or micro HDD) rather than keeping only recent streaming data allowable in memory, because data stream mining and tracking of past data stream are often required. In the general method using wavelets, a specific amount of streaming data from a sensor is periodically compressed into fixed size and the fixed amount of compressed data is stored into stable storage. However, differently from the general method, our method flexibly adjusts the compressing size based on a heuristic criterion. Experimental results with some real stream data show that wavelets technique is useful in data stream reduction and our flexible approach has lower estimation error than the general fixed approach.