Prediction of mustard yield using different machine learning techniques: a case study of Rajasthan, India

Area of Science:

• Agricultural Science
• Data Science
• Machine Learning

Background:

• Mustard is a vital oilseed crop in India, vulnerable to adverse weather conditions impacting yield.
• Effective crop yield prediction is essential for agricultural planning, storage, and trade decisions.
• Weather's influence on crop yield is complex, depending on variable magnitude and distribution patterns.

Purpose of the Study:

• To develop and compare machine learning models for accurate mustard crop yield prediction.
• To identify the most effective modeling approach for forecasting mustard yields based on weather data.

Main Methods:

• Developed six predictive models using long-term weather and mustard yield data.
• Employed variable selection (Stepwise Multiple Linear Regression - SMLR) and extraction (Principal Component Analysis - PCA) techniques.
• Utilized machine learning algorithms: Artificial Neural Network (ANN), Support Vector Machine (SVM), and Random Forest (RF).

Main Results:

• The Principal Component Analysis-Support Vector Machine (PCA-SVM) model demonstrated superior performance.
• PCA-SVM outperformed other individual models based on accuracy metrics (nRMSE, RMSE, RPD).
• Optimized combinations of models further improved prediction accuracy compared to individual models.

Conclusions:

• Machine learning, particularly PCA-SVM, offers a robust tool for accurate mustard yield prediction.
• Integrating weather data with advanced ML techniques enhances agricultural management and planning.
• Model optimization is key to maximizing prediction accuracy for critical crop forecasting.