Interpretable Time Series Predictions
by Eoin Delaney
Time series analysis may sound like a daunting task; however, anyone who is working with data that has a temporal component is a time series analysis! Some of the problems we are interested in tackling at VistaMilk include grass growth prediction and milk supply forecasting. Exploring trends and seasonal effects in the data can help us to predict future events, understand why past events occurred, and examine how these events could have been different.
The questions we are interested in include; How would calving earlier in the year and for a shorter period impact milk supply forecasts? Or how would our predictions about grass growth change if we had an exceptionally hot summer? Such questions pose two crucial challenges for researchers. Firstly, making models that can accurately predict events and secondly, understanding the model’s prediction in a way that it can be interpreted, trusted, and acted upon by end-users including farmers and industry partners.
An exciting area of research known as Explainable Artificial Intelligence (XAI) can help answer these questions. Research in VistaMilk enables collaborations between experts in diverse areas where combined knowledge plays a crucial role in predictive model development and deployment that will ultimately be of benefit to Irish farms both in terms of sustainability and food security.
Contact: Eoin Delaney (firstname.lastname@example.org)
Supervisor: Derek Greene (email@example.com) &
Mark Keane (firstname.lastname@example.org)