EUVIMEDEUVIMED entdecken
Crossref

Optimizing crop yield prediction in precision agriculture using random forest and particle swarm optimization: A data-driven approach to sustainable farming

Chen Jin

Sustainable Computing: Informatics and Systems · 2026-06 · 50 · 101336

DOI: 10.1016/j.suscom.2026.101336 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sustainable Computing: Informatics and Systems, 2026-06. Verfasst von Chen Jin. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 23 Literaturangaben. Gefördert von: Huangshan University.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Crop yield prediction stands as a pertinent factor when accounting for food securities and sustainable agriculture amidst climate variability and population pressure. Several regression models such as SVM, MLP, and CNN-RNN suffer from the drawback of being poorly generalized and inadequately hyperparameter-tuned, hence putting a limit on the performance and scalability of these models. With the goal of bypassing these challenges, this research puts forth the idea of a hybrid model wherein a Random Forest is combined with Particle Swarm Optimization to achieve higher prediction accuracies and model efficiencies. The uniqueness of this method rests on taking advantage of PSO to optimize hyperparameters dynamically and efficiently, thus while circumventing grid search, which is time-consuming, it also has the benefit of enhanced convergence speed. With RMSE = 0.20, MAE = 0.18, R2 = 0.98, and MAPE = 3.20%, the RF–PSO model has produced good results based on the Crop Yield Prediction Dataset, significantly outperforming the baseline techniques such as SVM (RMSE = 0.37), RFXG (0.22), and CNN-RNN (21.69 ± 3.10). Error distribution graphs, feature importance displays, and performance comparison charts all confirm the model's sturdiness with rainfall, temperature, and soil type identified as the chief predictive characteristics. Unlike other contemporary models, the RF–PSO guarantees high generalizability, low prediction errors, and practical utility in various agricultural settings. It is demonstrated by this study that RF–PSO is a scalable, interpretable, high-performance solution for precision agriculture and presents actionable insights to agronomists and decision-makers, while paving the way for prospective enhancements via multi-objective PSO, hybrid optimization, and explainable AI.

Förderung

  • Huangshan University

Zitieren

APA

Jin, C. (2026). Optimizing crop yield prediction in precision agriculture using random forest and particle swarm optimization: A data-driven approach to sustainable farming. Sustainable Computing: Informatics and Systems, 50, 101336. https://doi.org/10.1016/j.suscom.2026.101336

Vancouver

Jin C. Optimizing crop yield prediction in precision agriculture using random forest and particle swarm optimization: A data-driven approach to sustainable farming. Sustainable Computing: Informatics and Systems. 2026;50:101336. doi:10.1016/j.suscom.2026.101336

BibTeX
@article{Jin2026,
  author = {Jin, Chen},
  title = {Optimizing crop yield prediction in precision agriculture using random forest and particle swarm optimization: A data-driven approach to sustainable farming},
  journal = {Sustainable Computing: Informatics and Systems},
  year = {2026},
  volume = {50},
  pages = {101336},
  doi = {10.1016/j.suscom.2026.101336}
}
Kostenlos starten →