Performance Improvement of Regional Agricultural Forecasts with PECNET and State-Space Model

Mustafa Abdullah Hakkoz*, Serkan MacIt, Burak Berk Ustundag

*Bu çalışma için yazışmadan sorumlu yazar

Araştırma sonucu: ???type-name???Konferans katkısıbilirkişi

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 dilİngilizce
Ana bilgisayar yayını başlığı12th International Conference on Agro-Geoinformatics, Agro-Geoinformatics 2024
YayınlayanInstitute of Electrical and Electronics Engineers Inc.
ISBN (Elektronik)9798350380606
DOI'lar
Yayın durumuYayınlandı - 2024
Etkinlik12th International Conference on Agro-Geoinformatics, Agro-Geoinformatics 2024 - Novi Sad, Serbia
Süre: 15 Tem 202418 Tem 2024

Yayın serisi

Adı12th 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/BölgeSerbia
ŞehirNovi Sad
Periyot15/07/2418/07/24

Bibliyografik not

Publisher Copyright:
© 2024 IEEE.

Parmak izi

Performance Improvement of Regional Agricultural Forecasts with PECNET and State-Space Model' araştırma başlıklarına git. Birlikte benzersiz bir parmak izi oluştururlar.

Alıntı Yap