Predicting symptoms of downy mildew, powdery mildew, and gray mold diseases of grapevine through machine learning

Downy mildew, powdery and gray mold are major diseases of grapevine with a strong negative impact on fruit yield quality. These controlled by the application chemicals, which may cause undesirable effects environment human health. Thus, monitoring forecasting crop disease is essential to support integrated pest management (IPM) measures. In this study, two tree-based machine learning (ML) algorithms, random forest C5.0, were compared test their capability predict appearance symptoms diseases, considering meteorological conditions, spatial indices, number protection treatments frequency days in recorded previous year. Data collected Tuscany region (Italy), presence grapevine, from 2006 2017 divided an 80/20 proportion training set, data 2018 2019 tested as independent years for downy mildew mildew. The year cumulative precipitation April seven before day most important variables among those considered analysis predicting occurrence symptoms. best performance three was obtained algorithm C5.0 applying (i) technique deal imbalanced dataset (i.e., detected minority observations) (ii) optimized cut-off predictions. balanced accuracy achieved set 0.8 0.7 0.9 mold. models (2018 2019) lower accuracy, around both diseases. Machine able select predictors unravel complex relationships geographic bioclimatic