Open Collections

Medema, Emily

Agriculture is a vast industry that has the potential for fascinating data analysis and research. However, a lot of data collected by farms and researchers goes to waste as it is inaccessible or focused on monetary gain rather than the interests of researchers or farmers. In an age of big data and data-driven development, it is unfortunate that a majority of the data collected within agriculture is stored within local or proprietary files and thus a daunting analysis task. Ensuring this data is accessible and performing an analysis that is understandable would be a major step forward for the agricultural academia and industry environment. Filling this gap of open-sourced data analysis will allow for a more customized analysis to be performed on bovine data, widening the potential research avenues utilizing sensor data. We have transferred the data provided from CSV files to a DBMS hosted on the cloud, in this case, MySQL on a Digital Ocean Ubuntu Server through a generic, customizable script that reads and uploads all of the data. We then utilized Machine Learning models, specifically k-means, K Nearest Neighbours (KNN), and Isolation Forests, to detect outliers and inform our exploration into event detection. Through our transfer of data and analysis, we can see that having accessible data allows for the development of new detection models and a new understanding of the data itself.

Text

University of British Columbia. COSC 449

Vancouver : University of British Columbia Library

10.14288/1.0413675

Science, Irving K. Barber Faculty of (Okanagan); Computer Science, Mathematics, Physics and Statistics, Department of (Okanagan)

Undergraduate

DSpace