Özet
Accurately predicting the harvest time at regional scales is crucial for optimizing resource deployment and increasing agricultural productivity across various agricultural applications. Traditional methodologies, which have relied on sensor-collected climatic and environmental data such as Growing Degree Days, Vapor Pressure Deficit, Reference Evapotranspiration, Minimum Temperature, and Accumulated Rainfall, using regression models for smaller scale predictions, have not fully harnessed the potential of advanced time-series prediction algorithms for larger scales. This study introduces the Predictive Error Compensated Wavelet Neural Network (PECNET) algorithm, a sophisticated time-series prediction tool designed for a wide range of forecasting applications, including harvest time forecasting at the regional level, leveraging comprehensive data infrastructures. The research makes big steps forward in prediction accuracy and efficiency by combining PECNET with the Mealy machine model, which is a finite state automaton well-known for working with dynamic systems and state-dependent outputs. This integration enhances the model's ability to reflect the complex dynamics of phenological stages on prediction outcomes, leading to notable improvements in R2 scores: PECNET outperformed LSTM by 0.04 and Random Forest by 0.29 on the phenological sampling dataset, and by 0.08 over LSTM and 0.23 over RF on the uniform sampling dataset. The synergistic use of phenological stage sampling further boosts the performance of all models, underscoring PECNET's capability to deliver accurate, scalable, and context-sensitive predictions. This integrated approach not only advances agricultural forecasting but also sets a new benchmark for future research, demonstrating the value of combining cutting-edge machine learning techniques with established automata theory to enhance agricultural management practices.
Orijinal dil | İngilizce |
---|---|
Ana bilgisayar yayını başlığı | 12th International Conference on Agro-Geoinformatics, Agro-Geoinformatics 2024 |
Yayınlayan | Institute of Electrical and Electronics Engineers Inc. |
ISBN (Elektronik) | 9798350380606 |
DOI'lar | |
Yayın durumu | Yayınlandı - 2024 |
Etkinlik | 12th International Conference on Agro-Geoinformatics, Agro-Geoinformatics 2024 - Novi Sad, Serbia Süre: 15 Tem 2024 → 18 Tem 2024 |
Yayın serisi
Adı | 12th International Conference on Agro-Geoinformatics, Agro-Geoinformatics 2024 |
---|
???event.eventtypes.event.conference???
???event.eventtypes.event.conference??? | 12th International Conference on Agro-Geoinformatics, Agro-Geoinformatics 2024 |
---|---|
Ülke/Bölge | Serbia |
Şehir | Novi Sad |
Periyot | 15/07/24 → 18/07/24 |
Bibliyografik not
Publisher Copyright:© 2024 IEEE.