[M] Explaining Estrus Classifications

Master Assignment

Explaining estrus classifications

Type: Master CS

Location: Nedap @ Groenlo

Period: (TBD)

Student: (TBD)

If you are interested please contact :

Robin Ali (Nedap)

Introduction of company:

Nedap is a stock-listed technology company with roughly 700 employees making 191M annual revenue. The company focuses on people that create and uses sensing technologies in various sectors. The head quarter and main development site lies in Groenlo, the Netherlands. Livestockmanagement is one out of Nedap’s seven business units (BU). The BU creates technology to automate the tasks on a farm. This is important as farms around the world grow quickly in size and personnel is often untrained or unwilling to adhere to uncomfortable working schedules.

Project description

The Smart Tag is among Nedap’s flagship products, which registers a cow’s activity and behavior and sends the results to the cloud for analysis and recommendation. Detecting Estrus, or heat, of a cow is one of the tag’s core features. Currently, the detector is implemented as a tree-based classifier, trained by examples. In total the classifier combines roughly 50 features to decide whether a cow is currently in heat or not. In detail, the classification is done on the following features that are refreshed every two hours: 1) a so-called activity level, 2) the times for behavior such as eating and ruminating, 3) historical activity, and 4) general events (such as calving). The results are shown to the user in the form of a task list containing animals that have to be inspected. Due to the online characteristic of the data, the entry for every cow can therefore change once in two hours.

Although the classifier often provides satisfactory accuracy, it is sometimes difficult to explain to users why the classifier produced certain output, especially because both support employees and farmers often only have limited understanding of computer technology. A particular complication is the fact that the features change over time and new events can be registered all the time. 

The project’s task is to implement a state-of-the-art explainer and perform research into adaptations into the specific use-case.

Expected product:

Available resourses:

Nedap offers an innovative work environment, all the necessary data and infrastructure to perform such research. In particular, after signing a non-disclosure agreement, the student will get access to the production version of the classifier and close supervision with regular feedbacks.