Comparison between random forest and gradient boosting machine methods for predicting Listeria spp. prevalence in the environment of pastured poultry farms. (August 2019)

• Record Type: Journal Article
• Authors: Golden, Chase E.; Rothrock, Michael J.; Mishra, Abhinav
• Is Part Of: Food research international. Volume 122(2019)
• Page Start: 47
• Page End: 55
• Publication Date: 2019-08
• Subjects: Food safety -- Listeria spp. -- Alternative poultry production -- Boosted trees -- Random forest -- Gradient boosting machine -- Predictive microbiology; Food -- Analysis -- Periodicals; Food industry and trade -- Periodicals; Food industry and trade -- Canada -- Periodicals; Food Technology -- Periodicals; Food -- Periodicals; Food-Processing Industry -- Periodicals; Aliments -- Industrie et commerce -- Périodiques; Aliments -- Industrie et commerce -- Canada -- Périodiques; Aliments -- Recherche -- Périodiques; Food industry and trade; Canada; Periodicals; Electronic journals; 664.005
• DOI: 10.1016/j.foodres.2019.03.062
• Languages: English
• ISSNs: 0963-9969

Abstract

Abstract: Foodborne pathogens such as Listeria spp. contain the ability to survive and multiply in poultry farming environments, which provides a route of contamination for poultry processing environments and final poultry products. An understanding of the effect of meteorological variables on the prevalence of Listeria spp. in the farming environment is lacking. Soil and feces samples were collected from 11 pastured poultry farms from 2014 to 2017. Random forest (RF) and gradient boosting machine (GBM) predictive models were generated to describe and predict Listeria spp. prevalence in feces and soil samples based on meteorological factors at the farming location. This study attempted to demonstrate the use of GBM models in a food safety context and compare their use to RF models. Both feces models performed very well, with area under the curve (AUC) values of 0.905 and 0.855 for the RF and GBM models, respectively. The soil GBM model outperformed the RF model with AUCs of 0.873 and 0.700, respectively. The developed models can be used to predict the prevalence of Listeria spp. in pastured poultry farm environments and should be of great use to poultry farmers, producers, and risk managers. Graphical abstract: Unlabelled Image Highlights: Environmental samples were evaluated from poultry farms for Listeria prevalence. Listeria prevalence can be predicted by machine learning models based on weather. Wind speed, temperature, and humidity affect Listeria spp. prevalence. This study provides a framework for future use of machine learning in food safety.