Machine learning approaches are now highly valued in multiple areas of plant breeding, genomics, and disease identification. Our target is the building of a decision support system that can be used by stakeholders to predict the occurrence of mycotoxin
contamination of grain from a national and farm/field-level perspective. We will integrate the best practice critical control points as identified in Task 2, and key learnings from all other tasks to inform the decisions. Further, we will use a variety of irregularly structured,
spatio-temporally varying data sources to create predictions. The resulting models will be built into an easy to use dashboard which provides information to stakeholders via a decision support system (DSS).
Objectives
• Tailoring the full specification of the decision support system via input from partners and stakeholders as identified in task 2 and from
project learnings.
• Creating a pipeline for the final database into a format suitable for machine learning approaches.
• Feature engineering and exploratory data analysis to ensure data veracity.
• The exploration of appropriate IBM Auto-AI and bespoke machine learning algorithms to optimise predictive performance.
• The deployment and testing of the front end DSS via the IBM Cognos Dashboard system with a sample of stakeholders, with potential iterative performance updates
Lead Researcher, Institution & Other Institutions involved
Prof. Andrew Parnell (NUIM), all partners.