Files in this item



application/pdf9021672.pdf (8MB)Restricted to U of Illinois
(no description provided)PDF


Title:An electronic model for determining value and predicting sale price of Holstein cattle sold at public auction
Author(s):Dill, Dewayne Edward
Doctoral Committee Chair(s):Spahr, Sidney L.
Department / Program:Animal Sciences
Discipline:Animal Sciences
Degree Granting Institution:University of Illinois at Urbana-Champaign
Subject(s):Agriculture, Animal Culture and Nutrition
Economics, Agricultural
Abstract:A model was developed for determining the appraised value of Holstein dairy cattle using examples of 2648 heifers and cows professionally appraised prior to sale at public auction in 1987-1988. Data recorded in the sale catalog, additional information announced at the sale and type conformation scores were transformed into 137 concepts that described each animal sold. The importance of each concept was determined by calculating the correlation with appraised value and sale price. Independent factors for cows (24) and heifers (18) were generated using principal component analysis. Importance of each factor was determined by a modified stepwise regression analysis. The cluster component of a machine learning program, PLS (Probabilistic Learning System), was used to develop the model. The machine learning program generated a decision tree by repeatedly dividing the examples into subsets of increasingly similar examples. Multiple linear regression analysis was performed on examples in the terminal node. Concepts most related to appraised value were type conformation scores for cows and the dam's type classification score for heifers. Most significant heifer factors were (1) heifers with high type score and (2) embryo transfer heifers from high final type score dam with many ancestors over 11,364 kg milk yield, 455 kg fat yield and 90 points final type score. Most significant cow factors were (1) cows with high type score, (2) old cows with increasing production, (3) cows from high milk and fat record families, and (4) cows with low milk and fat yield from high milk and fat yield dams. Fifteen alternative models (5 each for cows, bred heifers and open heifers) were generated and tested using a test data set of 4 sales. The Spearman rank correlation of predicted value with actual sale price for each model tested ranged from 0.573 to 0.714 for cow models, 0.113 to 0.497 for open heifer models and 0.032 to 0.431 for bred heifer models.
Issue Date:1990
Rights Information:Copyright 1990 Dill, Dewayne Edward
Date Available in IDEALS:2011-05-07
Identifier in Online Catalog:AAI9021672
OCLC Identifier:(UMI)AAI9021672

This item appears in the following Collection(s)

Item Statistics