Document Type

Thesis - Open Access

Award Date

2023

Degree Name

Master of Science (MS)

Department / School

Agronomy, Horticulture, and Plant Science

First Advisor

Karl Glover

Keywords

Farinograph, Machine Learning, Random Forest, Spring Wheat, Stability, XG Boost

Abstract

Hard red spring wheat (Triticum aestivum L., HRSW) flour is mostly used to produce wheat-based foods where dough strength is a major quality component. Maintaining adequate levels of dough strength is a key objective in the development of HRSW cultivars. In commercial settings, dough strength is often measured using Farinograph stability. Due to resource constraints within breeding programs, flour quality is often measured by other methods, such as the Mixograph and Glutomatic. The objective of this research was to determine whether data from the Mixograph and Glutomatic could be used to properly predict Farinograph stability. Farinograph and Mixograph data spanning 6 years and 7 locations were used for the analysis. The Farinograph stability data provides a check for the accuracy of the predictive models, which were developed with the Mixograph and Glutomatic data. Two regression-based models were ran, using parameters obtained from the Mixograph and Glutomatic tests: Random Forest, and XGBoost. The Random Forest model had an average of 162 seconds difference between a predicted time and an observed time, while the XG Boost model had an average of 196 seconds difference between predicted and observed. Observations with low stability times often had predictions greater than their true times, while those with high stability times frequently had predictions which were less than their true times. Although this was the case, selecting materials in a breeding program with predictions of greater than 1200 seconds would most often lead to satisfactory results.

Publisher

South Dakota State University

Download

Share

COinS
 

Rights Statement

In Copyright