A new AI-based similarity measure for recommending optimal long-term crop rotation alternatives

Abstract

We present a tool that enables personalized recommendations for agricultural producers by offering long-term crop rotation alternatives that simultaneously optimize economic and environmental efficiency across multiple variables. To achieve this, we propose a new similarity metric between sequences of agronomic decisions, based on neural network techniques commonly used in the field of natural language processing. This metric allows us to select, from a set of Pareto-optimal solutions (generated by the AgrOptim simulation and optimization system), those sequences that are most similar to the typical and customary practices of each producer. In this way, we aim to reduce the barriers to adopting the recommendations generated by AgrOptim, facilitating the implementation of more sustainable and efficient cropping sequences and production decisions.