Computational intelligence techniques for decision making: with applications to the dairy industry

The dairy industry is a major resource in the Australian economy. In 1995-96 there were 13,888 dairy farms in Australia with 1.924 million dairy cows 2 and a total milk production of 8,716 million litres with a combined value of $AUD3 billion on the wholesale level. The industry’s success is dependent on increased animal productivity through breeding programs and efficient management. One of the main challenges in the industry therefore, is to improve the productivity of breeds through the design of efficient breeding programs. These programs aim to improve the genetic merit as well as the productivity of animals through identifying (selecting) and mating (allocating) animals with high genetic values, under the prevailing environmental and managerial conditions. To solve the selection and allocation problems, we need to predict the progeny’s expected productivity (ie. milk, fat, and protein yields) arising from any mating. From the previous discussion, the task of building a breeding program can be decomposed into three interrelated stages: prediction, selection, and allocation. In prediction, the expected merit of potential progeny is estimated from information collected and recorded about their parents and the environment. In selection, a set of sires and a set of dams are chosen for mating according to the overall goal of the program. In allocation, individual matings among the selected animals are decided according to the expected progeny merit as well as some preferences and goals. Reaching an optimal decision is only possible when both selection and allocation are solved simultaneously and in this case the problem is called mate-selection. The problem is hard when the planning is for a single generation to optimise a short-term goal. In long-term planning, additional objectives need to be considered such as the minimisation of inbreeding (ie. matings of related individuals). As a result, nonlinearity and conflicting objectives arise within the problem. Mate-selection decisions require progeny prediction and other information that can be retrieved from the databases currently kept by the farmers and their organisations for production and pedigree records. The objective of this thesis is to integrate Knowledge Discovery from Databases (KDD) with the Intelligent Decision Support System (IDSS) paradigm, to comprise what we call in this thesis Source: Australian Dairy Corporation ADC. Source: Australia Bureau of Statistics.