Neural network-assisted body weight prediction of goat kids using morphometric measurements in various growth phases

Body weight (BW) measurement of animals is an essential farm activity to monitor their growth and welfare but weighing equipment is often inaccessible for low-income farmers. This work utilized animal morphometrics to determine BW of Black Bengal goat kids using artificial neural networks. Over four months, 130 observations per growth phase (pre-weaning and post-weaning) were collected for morphometric measurements and BW. Twelve different body morphometric measures were taken, and correlated with BW. Various combinations of training algorithms (LM and GDX) with LOGSIG and TANSIG transfer functions were tested across 5 to 30 hidden layer neurons (HLNs). Results showed higher accuracy for BW prediction using heart girth and height at back (HAB) for the pre-weaning phase. However, the corpus length (CL) and HAB showed better model accuracy for post-weaning and combined growth phase. The selected ANN model showed superior results than the non-linear and linear regression models. ANN may therefore be used to predict BW of goat kids in place of other regression methods.