Machine Learning Application to Support Sow's Culling Decision
2010
Choe, Y.C., Seoul National University, Seoul, Republic of Korea | Lee, M.S., Jeonbuk Development Institute, Jeonju, Republic of Korea
The Machine Learning has been used widely from engineering to finance as a promising predictive analysis tool with large data sets. We apply the machine learning techniques to develop reference model to predict the sow's culling decision using pig farm data. We compared five machine learning techniques: logistic regression, decision tree, artificial neural network, k-nearest neighbor, and ensemble. The ensemble and logistic regression are well performed to predict the sow's culling in all parity. The ensemble is the best classifier for third to fifth parity and the logistic regression is the best for sixth and seventh. The neural network is excellent classifier for third, forth and seventh parity. The decision tree method is excellent classifier for fifth parity. The k-nearest neighbor performed poorly for all parity. The decision tree based on the CHAID is used to uncover the management strategies for farmers to replace their shows. It finds that sow's culling decision is based on the information including the number of weaned piglets per litter, piglets born alive per litter, and lactation days etc.
Mostrar más [+] Menos [-]Palabras clave de AGROVOC
Información bibliográfica
Este registro bibliográfico ha sido proporcionado por Korea Agricultural Science Digital Library