Assessing Transparency, Accuracy, and Consistency of Relative Importance of Cow-Calf Profitability Drivers Using Neural Networks versus Regression. (19th August 2020)

• Record Type: Journal Article
• Authors: Tester, Colson A.; Popp, Michael P.; Dixon, Bruce L.; Nalley, Lanier L.
• Is Part Of: Journal of agricultural and applied economics. Volume 52:Number 3(2020)
• Page Start: 352
• Page End: 367
• Publication Date: 2020-08-19
• Subjects: Artificial neural networks (ANNs), -- calving season, -- cow herd profitability, -- fertilizer use, -- herd size management, -- multivariate regression (MR); D24, -- Q12; Agriculture -- economics; Agriculture -- Economic aspects -- Periodicals; Agriculture -- Economic aspects -- Southern States -- Periodicals; Agriculture -- Economic aspects -- United States -- Periodicals; Agriculture -- Aspect économique -- Périodiques; Agriculture -- Aspect économique -- États-Unis -- Périodiques; Agrarökonomik; Agrarmarkt; USA; Welt; Agriculture -- Economic aspects; United States; Periodicals; 338.10975
• DOI: 10.1017/aae.2020.6
• Languages: English
• ISSNs: 1074-0708

Abstract

Abstract: Using both multivariate regression and artificial neural networks, the relative impact of variables affecting cow-calf profitability was examined over two cattle cycles for spring- and fall-calving herds that varied in size by time period analyzed when using different fertility management affecting forage yields with and without weather uncertainty. Neural networks had greater predictive accuracy than regression but at the cost of lesser transparency and predictive consistency. Explaining profitability, price, and quantity of cattle sold were consistently and respectively ranked first and second using both approaches. Importance rankings for hay sold and fertilizer were low and less consistent across techniques employed.