Data analysis provides a huge contribution to answering our motto question “How is the cow?”. Data from different sensor systems are combined, linked to further process data as well as historical data and are finally evaluated jointly. Hence, a forecast can be given about the animals’ health and welfare situation.
Background: In dairy farming, an increasing amount of data is generated from various sources: breed information, milk production data from milking technology, milk quality parameters from milk inspections, veterinary diagnoses, etc. All these data are usually maintained within herd management systems. Many new sensors and assistance systems are being developed recently that provide even more data. These lead to more complex evaluations, e. g. pedometers for activity measurements and heat detection, rumen boluses for rumen monitoring and tracking systems for animal positioning.
In order to use the data and extract meaningful information out of this ‘flood’, advancing developments in data analysis and prediction models are essential. Besides statistical methods, machine learning techniques and artificial intelligence are employed nowadays. In short, data analysis deals with answering our motto question “How is the cow?”. However, predicting the health status of a dairy herd is a complex task. There are huge differences between individual cows and their normal behaviour. Therefore, many training data is necessary to develop algorithms that aim to detect anomalies and decide whether it’s a heat, calving or disease event.
In experimental area 4, different analysis methods are examined and compared in order to achieve the most precise prognoses about the health and welfare situation of each individual of the dairy herd. This does not only answer the question about the current animals’ states
and provides action recommendations – e. g. via alert lists – but also enables the detection of deficiencies in herd management and the possibility to verify the success of the measures taken.