Event-Based Early Warning of Vineyard Disease Risk from Environmental Time Series

The paper proposes an event-based approach to predict transitions into vineyard disease‑risk periods within a 3–7 day window, rather than daily disease status. It defines events only after a minimum disease‑free gap to reduce label fragmentation and uses multi‑year agro‑meteorological data to capture humidity, rainfall, temperature, and seasonal patterns. Experiments with XGBoost, LSTM, and TCN models show that this formulation improves short‑horizon warning, highlighting trade‑offs among recall, lead time, and false alerts.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Machine Learning.