Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 268 - Extreme Machine Learning Methods and Applications: Domestic and International
Type: Topic Contributed
Date/Time: Tuesday, August 4, 2020 : 1:00 PM to 2:50 PM
Sponsor: Section on Statistical Consulting
Abstract #312857
Title: Empirical data-fusion approaches to generate model covariates
Author(s): Luca Sartore* and Jake Abernethy and Claire Boryan and Lu Chen and Kevin Hunt and CLIFFORD H SPIEGELMAN and Linda J Young
Companies: National Institute of Statistical Sciences and NASS and USDA and USDA and USDA and Texas A&M University and NASS
Keywords: Non-parametric; Machine Learning; Empirical estimates; Agricultural statistics; Artificial covariates
Abstract:

USDA's National Agricultural Statistics Service (NASS) publishes more than 500 reports every year including prospective planting estimates and planted acreage estimates for major crops at the state level. To produce these estimates, several surveys are conducted before and during the growing season. To enhance the accuracy of these estimates, NASS has investigated the use of a model that incorporates external sources of information such as soil type, weather and remote sensing data obtained at different temporal and spatial resolutions. Estimates are updated weekly as the input data evolve over time. Empirical approaches based on an ensemble of non-parametric techniques are investigated to generate a set of covariates by fusing together several sources of information. The resulting variables are then evaluated on two different tasks: the classification of prevented planting fields, and the performance in predicting crop at the field level. A case study using historical data in the state of Illinois is presented.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program