Imperial College London
Portrait Picture

ProfessorSamuelKounaves

Faculty of Engineering, Department of Earth Science & Engineering

Visiting Professor
 
 
 
//

Contact

 

+44 (0)7763 262 356s.kounaves Website CV

 
 
//

Location

 

2.34Royal School of MinesSouth Kensington Campus

//

Summary

 

Publications

Citation

BibTex format

@unpublished{Fang:2015,
author = {Fang, D and Oberlin, E and Ding, W and Kounaves, SP},
title = {A Common-Factor Approach for Multivariate Data Cleaning with an  Application to Mars Phoenix Mission Data},
url = {http://arxiv.org/abs/1510.01291v2},
year = {2015}
}
Download

RIS format (EndNote, RefMan)

TY  - UNPB
AB  - Data quality is fundamentally important to ensure the reliability of data forstakeholders to make decisions. In real world applications, such as scientificexploration of extreme environments, it is unrealistic to require raw datacollected to be perfect. As data miners, when it is infeasible to physicallyknow the why and the how in order to clean up the data, we propose to seek theintrinsic structure of the signal to identify the common factors ofmultivariate data. Using our new data driven learning method, the common-factordata cleaning approach, we address an interdisciplinary challenge onmultivariate data cleaning when complex external impacts appear to interferewith multiple data measurements. Existing data analyses typically process onesignal measurement at a time without considering the associations among allsignals. We analyze all signal measurements simultaneously to find the hiddencommon factors that drive all measurements to vary together, but not as aresult of the true data measurements. We use common factors to reduce thevariations in the data without changing the base mean level of the data toavoid altering the physical meaning.
AU  - Fang,D
AU  - Oberlin,E
AU  - Ding,W
AU  - Kounaves,SP
PY  - 2015///
TI  - A Common-Factor Approach for Multivariate Data Cleaning with an  Application to Mars Phoenix Mission Data
UR  - http://arxiv.org/abs/1510.01291v2
UR  - http://hdl.handle.net/10044/1/56736
ER  -
Download