Comparison Of Milk Production Parameters In Different Environment Predicted By Non-Linear Model And Artificial Neural Network

A mathematical formula describing the behavior of milk yield during a lactation referred to a lactation curve. Nowadays, artificial intelligence-based models, such as artificial neural networks (ANNs), are applied to predict milk curve, as well. The main advantages of ANNs are their ability to learn and control complex non-linear data. Dairy cattle breeding programs are based on milk production, which highly be affected from di↵erent climate situations. To predict the effect of temperature-humidity-index (THI) on lactation curve and compare the ANN and some di↵erent nonlinear functions (Wood, Wilmink) a dataset of ۱۰۷۹۱۴ milk test days from first parity cows in ۷ provinces located in humid and arid area during ۲۰۰۰ to ۲۰۱۶ was used. THI index categorized to ۵ stress classes (<۵۲, ۵۲-۶۱, ۶۲-۷۱, ۷۲-۸۲, >۸۲). ANN model was developed according to ۱-Multilayer Perceptron with ۵ nodes in layer, Sigmoid Axon function and Levenberg Marquadt learning rule by NeuroSolution V.۵ software. Nonlinear models were analyzed by NLIN procedure of SAS V.۹.۴. The goodness of fitted models and comparison determined by using the coe cient of determination (R۲) and residual mean square (MSE). The fixed effects included record number of lactation, season of milk record, age of cow (days), herd, year of calving, stress classes, number of milking. The results revealed that ANN had the highest accuracy (R۲=۰.۴۹۳) and the lowest error (RMSE=۰.۰۸۹), described the milk curve better than the Wilmink (R۲= ۰.۴۰۲, RMSE= ۶.۲۸۹) and Wood (R۲=۰.۳۸۸, RMSE= ۰.۲۳۲). The best condition for milk production was up to THI=۷۱ and thereafter it was decreased. According to (Lees et al,۲۰۱۹), the maximum daily air temperature has a great e↵ect on reducing milk yield. During warming, to a critical range of THI of ۷۰–۷۲, the performance of dairy cattle is inhibited, and deterioration in milk yield occurs at THI from ۷۲ to ۷۸, The results of this research are consistent with the current research.Therefore, ANN model is recommended for describing the milk production curve of Iranian Holstein cattle

کلیدواژه ها:

ANN ، THI ، Wilmink ، Prediction