Agricultural Knowledge Discovery Based on Interpretable Machine Learning and Visual Analytics

The goal of this research is to implement Artificial Intelligence (AI) to automatically extract and visualize critical values in agricultural knowledge. Many environmental factors such as climate, soil, light, etc., influence crop production, but not much is known about how much each factor influences crop growth, and what the most suitable value is for each factor for the crop. Farmers have long been concerned about these issues, but if we can master the knowledge of planting, we will be able to cultivate more accurately, enhance production, and improve planting methods. As an example, this study uses the Eustoma grandiflorum farm in Changhua, Taiwan, to show the visual analytics of the major agricultural knowledge mining process, which includes the three primary visual analytics: critical value of light intensity, growth prediction, and growth stimulation. The “critical value of light intensity” helps farmers quickly determine the precise thresholds for the impact of variable values on crops. The “growth prediction” helps farmers in predicting future growth using current planting conditions, and the “growth simulation” helps them in examining the impact of different factors on crop growth in real-time interaction and recommending specific Eustoma grandiflorum light-emitting diodes (LED) lighting knowledge based on the visualization results. This system integrates several visualization tools that can be used to conduct in-depth analyses of certain crop development parameters in a multi-faceted and efficient manner.