235,389 Pages

## Herd monitoring system (DRGTTST6.NET)

By combining several copies of the 12-sow networks in figure 3 a system for herd monitoring can be build. We will use a modification of the technique of ?REFERENCE

A simplified version of the network is shown in figure . A node with herd level is defined for each week. The herd level node at week $t+1$ is a child of the herd level node at week $t$. The transition matrix is defined such that the probability of transition from state $i$ to state $i+1$ is 0.85, while the transition probability from state $i$ to each of the remaining 6 states $j$, where $i \neq j$, is $(1 - 0.85)/6 = 0.025.$

Figure 4: Network for weekly monitoring of pregnancy rate

The present network only covers three weeks, but the network can easily be extended to a much longer period. Because of the underlying algorithm the complexity of the calculations only grows linearly with each additional week. (This can be studied closer, if the Information topic is selected in the Hugin programme).

This model readily makes it possible to include the information from the different sources (heat detection, pregnancy testing and farrowing) into an updated prediction of the herd level. The present network only monitors the herd level of pregnancy rate, while the quality of heat detection and pregnancy test is assumed constant. This can easily be modified as well, to allow monitoring of these herd parameters.