Irrigação Inteligente Sensível à Espécie baseada em Deep Learning e Dados Ecológicos Globais
Irrigação Inteligente Sensível à Espécie baseada em Deep Learning e Dados Ecológicos Globais
-
DOI: https://doi.org/10.22533/at.ed.441122601041
-
Palavras-chave: ..
-
Keywords: ..
-
Abstract: This work presents a species-sensitive smart irrigation approach ba- sed on the integration of plant ecological attributes and machine learning mo- dels, aiming at more efficient water use in agricultural systems. A system was developed in Python with a web interface to estimate daily water demand from environmental variables, phenological stage, and species ecological characte- ristics. The experimental dataset consisted of 2,800 synthetic samples generated from botanical and climatic profiles. Algorithms such as Ridge regression, KNN, and neural networks were evaluated using metrics including MAE, RMSE, and R2. The results showed that neural models achieved better performance, with the shallow MLP obtaining an RMSE of 0.2338 and an R2 of 0.9295 on the test set. The findings suggest that combining ecological data with machine lear- ning is promising for representing species-specific water requirements and sup- porting more adaptive irrigation systems. Although synthetic data were used, the approach demonstrated consistency and potential for application with real- world data.
- Gabrielly de Queiroz Pereira
- Marcos Monteiro Junior
- Rodrigo Adamshuk Silva