MODELING AND ALGORITHM DEVELOPMENT FOR RED SHEEP
Abstract
Author(s): Bhupendra Sharma and Dr. Subhash Chandra Sikdar
Modeling animal growth and production data can be helpful in understanding the pattern of growth and production for different species of animals. Further this can help in evolving strategies to addres s low productivity and augment production as well. Analysis of longitudinal growth data of Madhya Pradesh red sheep using Linear Mixed Effects (LME) model indicates that the conditional quadratic model with heterogeneous AR (1) error covariance structure is good with the covariates, dam’s weight at lambing and gender of lamb showing marked influence on the growth parameters. Season of birth is found to be significant only for growth rate and not for the intercept. Among the ANN models developed for the longitudinal growth data of Madhya Pradesh red sheep, ANN (Artificial Neural Effects) model using Multilayer Perceptron (MLP) architecture is found to perform better when compared to the ANN model based on RBF architecture. Further, when the ANN model based on MLP architecture is compared with the best LME model, it is found that ANN model based on MLP architecture is better than the LME model based on the reliability coefficient concluding that ANN can be considered as an useful alternative modeling technique for longitudinal growth data of Madhya Pradesh red sheep with better predictive ability.
