Improved SVM-Based Soil-Moisture-Content Prediction Model for Tea Plantation.

Area of Science:

• Agricultural Science
• Environmental Science
• Data Science

Background:

• Accurate soil moisture content (SMC) prediction is vital for tea plantation irrigation and productivity.
• Traditional SMC prediction methods are costly and labor-intensive.
• Existing machine learning models for SMC prediction often suffer from insufficient data.

Purpose of the Study:

• To develop an improved support-vector-machine (SVM) based model for accurate soil moisture content (SMC) prediction in tea plantations.
• To address limitations of existing methods by incorporating novel features and optimizing SVM performance.
• To enhance predictive performance, especially when limited real data is available.

Main Methods:

• Developed an enhanced SVM model incorporating novel features for SMC prediction.
• Optimized SVM hyper-parameters using the Bald Eagle Search (BES) algorithm.
• Utilized a comprehensive dataset including soil moisture measurements and environmental variables (rainfall, temperature, humidity, soil type).
• Applied feature selection techniques to identify the most informative variables.

Main Results:

• The improved SVM model demonstrated superior performance in predicting soil moisture content compared to traditional SVM and other machine learning algorithms.
• Achieved high accuracy with R² of 0.9435, MSE of 0.0194, and RMSE of 0.1392.
• The model showed robustness and generalization capabilities across different time periods and locations.

Conclusions:

• The proposed SVM-based model provides timely and accurate soil moisture predictions for tea plantations.
• Enables informed irrigation scheduling and water resource management, enhancing tea crop yield.
• Minimizes water usage and reduces environmental impact through optimized irrigation practices.