Machine learning‐based screening of pesticide adjuvants for reduction of pesticide losses on wheat leaf surfaces

Abstract Using pesticide spray adjuvants to improve the physicochemical properties of pesticide liquids can effectively increase droplet deposition on target surfaces. However, there are no clear guidelines for the selection among complex and diverse adjuvants. We aimed to build a model for screening pesticide adjuvants through machine learning. In this paper, five machine learning classification models (decision tree, support vector machine, random forest, logistic regression, and naive Bayes) were developed to predict droplet deposition performance on the superhydrophobic plant leaf surfaces based on the physicochemical properties of pesticide liquids. Among these models, decision tree, support vector machine, and random forest exhibited superior performance. Significance analysis showed that adhesion, equilibrium surface tension, and contact angle are critical factors influencing droplet deposition on wheat leaves. Notably, the decision tree model, due to its simplicity and intuitiveness, is particularly suitable for field applications. Our findings provide a platform for the rapid and accurate screening of pesticide spray adjuvants by simply measuring the physicochemical properties of pesticide droplets.