Practical aspects for classification of performance parameters of pigs by decision trees
2013
Schlegel, H., Anhalt University of Applied Sciences, Bernburg (Germany)
The potential of performance of breeding sows and other livestock gets described by a variety of data. Those data get analysed constantly. Aims of investigations are defined and approved. Therefore different methods are used, well known ones and relatively new ones. With the help of those different methods relationships should be represented and interpreted. One in comparison - relatively new way of Data Mining is the classification with the help of decision trees. There is one big advantage, because decision trees do not have any conditions according to the scale and distribution of parameters. Many factors can get analysed at the same time. Different kinds of decision trees can be used for known and unknown interactions between parameters. That is also an advantage. For representation of the usage of decision trees in livestock production one practical example has been chosen. The main aim of this investigation is the interpretation of the results from a biological and economical point of view. In addition to that the way of using a decision tree should get demonstrated. In this case, 1.780 sows of one pig breeding farm are used. They got classified according to the parameter of number of total born piglets, live born piglets, weaned piglets and number of litter, sire and technician for insemination. It is very important to derive the correct sequence steps from the statistical results. That's necessary for all interpretations.
显示更多 [+] 显示较少 [-]