Event Title

Problems Related to Missing Data in Hydrologic Modeling: Implications and Solution

Location

Eccles Conference Center

Event Website

http://water.usu.edu/

Start Date

3-28-2006 1:40 PM

End Date

3-28-2006 2:00 PM

Description

A common practice in pre-processing data for hydrological modeling is to ignore observations with any missing variable values at any given time step, even if it is only one of the independent variables that is missing. These rows of data are labeled incomplete and would not be used in either model building or subsequent testing and verification steps. This is not necessarily the best way of doing it as information is lost when incomplete rows of data are discarded. Learning algorithms are affected by such problems more than physically-based models as they rely heavily on the data to learn the underlying input/output relationships. In this study, the extent of damage to the performance of the learning algorithm due to missing data is explored in a field-scale application. We have tested and compared the performance of two well-known learning algorithms, namely Artificial Neural Networks (ANNs) and Support Vector Machines (SVMs) for short-term prediction of groundwater levels in a well field. A comparison of these two algorithms is made using various percentages of missing data. In addition to understanding the relative performance of these algorithms in dealing with missing data, a solution in the form of an imputation methodology is proposed for filling the data gaps. The proposed imputation methodology is tested against observed data.

This document is currently not available here.

Share

COinS
 
Mar 28th, 1:40 PM Mar 28th, 2:00 PM

Problems Related to Missing Data in Hydrologic Modeling: Implications and Solution

Eccles Conference Center

A common practice in pre-processing data for hydrological modeling is to ignore observations with any missing variable values at any given time step, even if it is only one of the independent variables that is missing. These rows of data are labeled incomplete and would not be used in either model building or subsequent testing and verification steps. This is not necessarily the best way of doing it as information is lost when incomplete rows of data are discarded. Learning algorithms are affected by such problems more than physically-based models as they rely heavily on the data to learn the underlying input/output relationships. In this study, the extent of damage to the performance of the learning algorithm due to missing data is explored in a field-scale application. We have tested and compared the performance of two well-known learning algorithms, namely Artificial Neural Networks (ANNs) and Support Vector Machines (SVMs) for short-term prediction of groundwater levels in a well field. A comparison of these two algorithms is made using various percentages of missing data. In addition to understanding the relative performance of these algorithms in dealing with missing data, a solution in the form of an imputation methodology is proposed for filling the data gaps. The proposed imputation methodology is tested against observed data.

https://digitalcommons.usu.edu/runoff/2006/AllAbstracts/23