Date of Award:

8-2026

Document Type:

Thesis

Degree Name:

Master of Science (MS)

Department:

Computer Science

Committee Chair(s)

Yang Chi (Committee Chair), Stephan van Vliet (Committee Co-Chair)

Committee

Yang Chi

Committee

Stephan van Vliet

Committee

Mahdi Nasrullah Al-Ameen

Abstract

Beef is one of the most nutrient-rich foods in the human diet, providing high-quality protein, iron, omega-3 fatty acids, B vitamins, and a wide range of other compounds important to health. However, current nutrition scoring systems used on food labels were designed to compare different foods to one another — for example, beef versus broccoli — and do not work well for judging the nutritional quality of different beef samples relative to each other. A grass-fed steak and a conventionally-finished steak can carry nearly identical Nutrition Facts panels while differing substantially in their content of omega-3 fatty acids, vitamins, and plant-derived antioxidant compounds that cattle absorb during pasture grazing.

This thesis develops a new scoring system specifically for beef and applies it to 377 samples analyzed for 30 different compounds. The system learns which compounds are most important for distinguishing beef samples directly from the data itself rather than assuming these weights in advance. When applied to the sample dataset, the system distinguishes grass-fed and grain-fed samples with approximately 89 percent accuracy. It also identifies six compounds — iron, omega-3, protein, coenzyme Q10, vitamin B5, and vitamin B6 — as the most important for characterizing beef nutritional quality, alongside a family of plant-derived antioxidant compounds that reflect whether the cattle were raised on pasture.

The system can also identify samples whose measured compound profile does not match the production label supplied by the vendor — for example, a sample labeled as grass-fed whose compound profile is actually more consistent with grain-finished beef. This provides a potential tool for supply-chain authentication that could complement existing methods used by regulators and third-party certifiers. The findings offer a working example of how modern multi-nutrient food composition data, analyzed with statistical tools capable of resolving high-dimensional structure, can be used to score, differentiate, and authenticate individual samples of a single food category.

Download

Share

COinS

Copyright for this work is retained by the student. If you have any questions regarding the inclusion of this work in the Digital Commons, please email us at .