Keywords
data mining, independent sources, gaussian distribution, spatio-temporal data, monitoring
Start Date
1-7-2008 12:00 AM
Abstract
The current work deals with application of independent virtual sources as a toolfor analysis and storing groundwater monitoring data. Estimation of virtual sources isbased on learning from data principle. Virtual sources allow to analyse the main featuresof the process. They can be useful for both data exploratory analysis and prediction. Thiswork considers two methods for independent virtual sources construction: (1) based onusing Central Limit Theorem and (2) based on the mixture of Gaussians. The methodswere applied to real spatio-temporal data on groundwater level dynamics (2D spatial case)and groundwater contamination by radioactive nitrates (3D spatial case). These data setsare characterized by heterogeneous sample distribution both in space and time.
Virtual Sources for Spatio-temporal Monitoring Data Analysis
The current work deals with application of independent virtual sources as a toolfor analysis and storing groundwater monitoring data. Estimation of virtual sources isbased on learning from data principle. Virtual sources allow to analyse the main featuresof the process. They can be useful for both data exploratory analysis and prediction. Thiswork considers two methods for independent virtual sources construction: (1) based onusing Central Limit Theorem and (2) based on the mixture of Gaussians. The methodswere applied to real spatio-temporal data on groundwater level dynamics (2D spatial case)and groundwater contamination by radioactive nitrates (3D spatial case). These data setsare characterized by heterogeneous sample distribution both in space and time.
