Intelligent irrigation strategy model for farmland using dung beetle optimization-random forest algorithms

The development of innovative agricultural management technologies is very urgent for global climate change and water scarcity. Intelligent irrigation technology has emerged as a precise management tool, which could enhance crop yield, quality and conserve water resources. This study proposes an optimized machine learning prediction model using the Dung Beetle Optimization-Random Forest (DBO-RF) algorithm, thus improving irrigation predictability. The model integrates multi-source data, including time-series features, agricultural meteorological data, and irrigation management specifics, precise hyperparameter tuning could be performed via the sequence decomposition-based Dung Beetle Optimization (DBO) algorithm. Field experiments were conducted in the unmanned rice fields at the Zhuanghang Experimental Station in Fengxian District, Shanghai, China. Essential meteorological and irrigation data were collected systematically. The obtained results demonstrated that the DBO algorithm significantly could enhance the Random Forest (RF) model's predictive accuracy, the Mean Absolute Error (MAE) and the Mean Square Error (MSE) were reduced to 0.30321 and 0.16382 respectively, the coefficient of determination (R²) had increased to 0.86255. Testing across diverse datasets revealed the DBO-RF model possesses robust generalization capability and consistently high predictive performance. This research provides valuable insights into agricultural meteorological data analysis and irrigation management, particularly for complex, multi-source data. The developed intelligent irrigation system dynamically adapts to environmental changes, optimizing water resource utilization and improving crop yields through an adaptive, real-time irrigation strategy.