Skip to main content
ARS Home » Southeast Area » Dawson, Georgia » National Peanut Research Laboratory » Research » Publications at this Location » Publication #82383
Title: ESTIMATION OF AFLATOXIN CONTAMINATION IN PREHARVEST PEANUTS USING NEURAL NETWORKS

Author
item PARMAR, R - UNIVERSITY OF GEORGIA
item MCLENDON, R - UNIVERSITY OF GEORGIA
item HOOGENBOOM, G - UNIVERSITY OF GEORGIA
item Blankenship, Paul
item Cole, Richard
item Dorner, Joe

Submitted to: Transactions of the ASAE
Publication Type: Proceedings
Publication Acceptance Date: 5/31/1997
Publication Date: N/A
Citation: N/A

Interpretive Summary: Interpretive summary not necessary for proceedings.

Technical Abstract: The objectives of this study were to examine the variables that affect the aflatoxin contamination process and to develop a model to estimate contamination levels. Artificial neural networks and linear regression models were used to model the contamination levels. Seven years of preharvest peanut aflatoxin data were used to develop and evaluate the models. The data were randomly divided into a training set and a test set for the model. The inputs considered were: soil temperature, drought duration, crop age, and accumulated heat units. The accumulated heat units were computed based on threshold soil temperatures ranging from 23 to 29 C. The R2-values for the training and the test sets were 0.9250 and 0.9522, respectively. Stepwise linear regression was also applied to develop a regression model for estimating aflatoxin values. The highest R2-values of 0.822 and 0.809 for the training and test sets, respectively, were achieved with the regression model when all four variables were selected as input factors and accumulated heat units were computed using a threshold temperature of 29 C. This study showed that artificial neural networks can be used to estimate aflatoxin contamination in peanuts. The artificial neural networks also performed better than traditional stepwise linear regression techniques.