E de la Garza Institute for Goat Research Langston University
Workshops & Field Day Newsletter Newsletter Subscription Demonstrations Demonstrations Langston University Research Building
Goat Menu




Classification tree analysis of grazing behavior in goats

T. A. Gipson, A. R. Askar, A. Beker, R. Puchala, A. Asmare, G. D. Detweiler, and A. L. Goetsch

Journal of Animal Science 86(E-Supplement 2):99-94. 2008

Electronic monitoring equipment may allow for characterization of grazing behavior without potential effects of human visual observation. Translating equipment output into specific activities, however, is challenging. Therefore, this study was conducted to develop means of predicting grazing behavior based on visual observation from output of currently available electronic monitoring systems. There were 1,538 5-min observations of grazing activity (G = grazing; RL = resting, lying; RS = resting, standing; W = walking) at two locations collected by four observers on 28 goats over 4-d periods. There were 390, 627, 478, and 43 observations for G, RL, RS, and W, respectively. Goats were fitted with GPS collars (GPS 3300, Lotek, Newmarket, Ontario, Canada) to ascertain distance between consecutive GPS fixes. Collars were equipped with left-right (X-activity), forward-backward (Y-activity), and head-down motion sensors. A leg activity/position sensing system (IceTag, IceRobotics, Midlothian, Scotland, UK) was employed to determine stepping, standing, and lying. Classification tree analysis was conducted using CARTŪ software. A decision tree, which is a diagram representing a classification system, with a minimum relative cost criterion of 0.560 yielded 18 terminal nodes. Prediction success rate for G was 70.3% (i.e., 274, 35, 48, and 33 G observations were classified into G, RL, RS, and W terminal nodes, respectively). Success rate for RL was 74.0% (57, 87, and 19 RL observations classified as G, RS, and W, respectively). Success rate for RS was 48.5% (93, 106, and 47 RS observations classified as G, RL, and W, respectively). Success rate for W was 83.7% (5, 1, and 1 W observations classified as G, RL, and RS, respectively). Output from currently available electronic monitoring equipment systems can be used to predict grazing behavior of goats based on visual observation; however, prediction success rate is less than optimal. Other potential monitoring equipment should be evaluated to improve success rate.


Extension Activities   |   Research Activities   |   Other Activities
Library Activities   |   Quiz   |   Search   |   About Us   |   Contact Us   |   Faculty & Staff
Research Extension Home   |   Top of Page

Copyright© 2000 Langston University   • Agricultural Research and Extension Programs
P.O. Box 730  • Langston, OK  73050 • Phone 405.466.3836