Machine Learning Extreme Values of Insect Infestations in Hop Fields
keywords MACHINE LEARNING
Reference persons GIOVANNI SQUILLERO
External reference persons Dr. Alberto Tonda
Research Groups GR-05 - ELECTRONIC CAD & RELIABILITY GROUP - CAD
Thesis type CODING AND EXPERIMENTAL, MODELING AND DATA ANALYSIS
Description Climate change is influencing agriculture in different ways, ranging from extreme draughts to alterations of the life cycle of parasite insects. Being able to predict the moment where the larvae of parasites are going to hatch would be invaluable to farmers, as it would allow them to spray pesticides only when needed, reducing costs and environmental pollution. However, readily available data on the subject is scarce, and the machine learning of rare events, or, in other terms, learning the extreme values of a distribution, is a hard task.
Starting from real-world data collected over the course of 18 years by the Slovenian Institute of Hop Research and Brewing, the objective of this internship is to use machine learning approaches to predict the peak appearance of the European corn borer (ostrinia nubilalis), a parasite of hop and corn fields. The candidate will explore several techniques, ranging from classical regression to learning of extreme values, with the main objectives of gaining a better understanding of the problem and ultimately obtaining reliable predictions.
The ideal candidate should feel comfortable programming in Python, and be passionate about data science and machine learning. Previous experience with modules for machine learning or data science, such as scikit-learn, TensorFlow, pandas, or keras, is a big plus, but not strictly necessary.
Deadline 22/05/2020 PROPONI LA TUA CANDIDATURA