Decision support in breeding program in the Kenya Dairy cattle Prediction and Rule extraction using Neural Network System
Abstract
An important characteristic of all natural systems is the ability to acquire knowledge through
experience and to adapt to new situations. This report presents findings on the use of neural
networks to make predictions on the performance of daughter first lactation milk yield in Holstein
Friesian cattle. Such prediction would ultimately lead to defining optimal breeding strategies.
Data consisting of 6095 lactation records made by cows from 76 officially milk recorded Holstein
Friesian herds and collected between 1988 and 2005 were used to predict the performance of the
offspring based on some of the genetic traits of their parents using neural networks which was
compared with linear regression model technique as baseline. The data was sorted using the
current criteria for selection and a univariate animal model with relationships and the Derivative
Free Restricted Maximum Likelihood procedure was used to predict the sire breeding values for
milk yield.
More accurate predictions were obtained by a neural network with 1 hidden unit than linear
regression as can be seen by comparing the correlation coefficients and root mean square errors.
This suggests that a non-linear relationship among the feature variables exists in the data and that
these are learned by the hidden layer of the neural network. Feature sets which included sire
information had high correlation coefficient.
The black-box nature of neural networks was explained by extracting rules with domain expert and
autoclass for both the continuous and the discrete valued inputs. Rules for discrete valued inputs
for both categorization performed better on the 'low' and 'high' levels. This implied that
performance at the two extremes are more important than average performance and that the client
is particularly concerned with identifying mating with good potential and avoid matings with poor
potential
Publisher
School of Computing and Informatics
Description
MSc