|
Activity Number:
|
477
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 7, 2008 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statisticians in Defense and National Security
|
| Abstract - #301511 |
|
Title:
|
Piecewise Linear Approximation of Data with Removable or Jump Discontinuities
|
|
Author(s):
|
Patricia H. Carter*+
|
|
Companies:
|
Naval Surface Warfare Center
|
|
Address:
|
18444 Frontage Rd, Dahlgren VA, 22448-5161,
|
|
Keywords:
|
piecewise linear approximation ; time series ; compression ; segmentation ; discontinuous data ; data reduction
|
|
Abstract:
|
There are many algorithms for the compression of time series data by approximation using piecewise linear functions. As usual the approximation trade off is compression rate versus a global measure of the error as well as the resulting errors in determining size and locations of events of interest. Here we introduce an agglomerative approximation algorithm designed for time series data that is well-modeled by functions with removable or jump discontinuities. This algorithm is exact on appropriate piecewise linear models with isolated removable and jump discontinuities, and outperforms generic bottom-up and random iteration methods at similar computation cost and algorithmic complexity.
|
