This work develops self-training and adaptive prediction models to enhance Decision Support Systems (DSS) in precision agriculture. The models optimize the prediction of agronomic and phytosanitary risks through the integration of Artificial Intelligence (AI) and hierarchical Bayesian statistical approaches, using RStan. Two models were created: one for durum wheat, based on a Transformer architecture to analyze weather data and satellite imagery and predict the NDVI, and another for grapevine, employing a hierarchical Bayesian approach that integrates data from sensors and traditional DSS to estimate stress risk. Tested in Sardinia, the models demonstrated strong predictive capabilities, enabling more timely interventions. Self-training, the integration of heterogeneous data sources, and adaptability to local context enhance the effectiveness of DSS, contributing to the reduction of phytopharmaceutical use and promoting sustainable agricultural practices."
Questo lavoro sviluppa modelli previsionali auto-addestrabili e adattabili per migliorare i Sistemi di Supporto Decisionale (DSS) in agricoltura di precisione, ottimizzando la previsione dei rischi agronomici e fitosanitari tramite Intelligenza Artificiale (AI) e approcci statistici gerarchici bayesiani usando RStan. Sono stati realizzati due modelli: uno per il frumento duro, basato su un Transformer per analizzare dati meteo e immagini satellitari e prevedere l’NDVI, e uno per la vite, con un approccio bayesiano gerarchico che integra dati da sensori e DSS tradizionali per stimare il rischio di stress. Testati in Sardegna, i modelli hanno mostrato una valida capacità predittiva, consentendo interventi più tempestivi. L’auto-addestramento, l’integrazione di dati eterogenei e l’adattabilità al contesto locale migliorano l’efficacia dei DSS, contribuendo a ridurre l’uso di fitofarmaci e promuovere pratiche agricole sostenibili.
Modelli di previsione del rischio agronomico fitosanitario auto addestrabili e adattabili al contesto locale
GOSAMO, EMANUELE
2025
Abstract
This work develops self-training and adaptive prediction models to enhance Decision Support Systems (DSS) in precision agriculture. The models optimize the prediction of agronomic and phytosanitary risks through the integration of Artificial Intelligence (AI) and hierarchical Bayesian statistical approaches, using RStan. Two models were created: one for durum wheat, based on a Transformer architecture to analyze weather data and satellite imagery and predict the NDVI, and another for grapevine, employing a hierarchical Bayesian approach that integrates data from sensors and traditional DSS to estimate stress risk. Tested in Sardinia, the models demonstrated strong predictive capabilities, enabling more timely interventions. Self-training, the integration of heterogeneous data sources, and adaptability to local context enhance the effectiveness of DSS, contributing to the reduction of phytopharmaceutical use and promoting sustainable agricultural practices."| File | Dimensione | Formato | |
|---|---|---|---|
|
PDFA_TESI_GOSAMO_EFFrontePON37_MR_lf.pdf
accesso aperto
Licenza:
Tutti i diritti riservati
Dimensione
10.92 MB
Formato
Adobe PDF
|
10.92 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/223301
URN:NBN:IT:UNITS-223301