Abstract
Assessing animal response during production stage is a time-consuming task, and new technologies are necessary to improve accuracy. Most variables used to estimate the animal status include the rearing environment (dry bulb temperature and relative humidity), physiological indicators (body temperature, cortisol level, respiratory frequency, and immunological response), and various group behaviors (fights, social interactions, and vocalization). Observation and recording of the majority of these variables on the farm are challenging and time-consuming, and therefore an automatic way of assessing the animal needs is highly desirable. The machine learning focus is used on a prediction; without necessarily assuming a functional distribution for the data. Several solutions have been developed such as a classifying algorithm for detecting pig screams with good results (72% of sensibility, 91% of specificity, and 84% accuracy). One way of envisioning the problem of learning is to imagine a search through a space of possible concept descriptions for one that fits the data. The statistical approach looks within a set of data and try to predict the whole. Machine learning deals with the entire set of data trying to understand the whole. Decision tree algorithms are capable of handling continuous-valued attributes. The method is intuitive and straightforward predictive models, making them an adequate selection when decision rules are essential. It is possible to build up solution for the classification of the activity status of an animal by asking a series of simple yes/no questions. The best fit is that which leads to the most homogeneous groups. In the presentation, several examples of algorithms are presented for helping decision-making in intensive animal production. Key Words: