Optimizing potato yield predictions in Uttar Pradesh, India: a comparative analysis of machine learning models

Area of Science:

• Agricultural Science
• Machine Learning Applications
• Data Science in Agriculture

Background:

• Potatoes are a crucial staple food for global food security and poverty reduction.
• India is a leading global potato producer, making accurate yield forecasting essential.
• Sustainable agriculture and food supply chain management require precise yield predictions.

Purpose of the Study:

• To compare the performance of five machine learning models for potato yield forecasting.
• To identify the most effective machine learning approach for predicting potato yields in Uttar Pradesh, India.
• To provide accurate forecasts for proactive planning in food supply and resource optimization.

Main Methods:

• Collected 16 years (2005-2021) of time series data on potato yields and weather variables for seven districts in Uttar Pradesh.
• Applied data detrending and weather index processing.
• Trained and validated five machine learning models: Elastic Net (ELNET), Random Forest, Artificial Neural Network (ANN), Extreme Gradient Boosting (XGBoost), and Support Vector Regression (SVR) using 70% training and 30% testing data splits.

Main Results:

• The Artificial Neural Network (ANN) model demonstrated superior performance, achieving the highest R² values and lowest error metrics.
• The performance ranking of the models was: ANN > XGBoost > Random Forest > ELNET > SVR.
• The ANN model achieved over 98% accuracy in predicting future potato yields in the studied districts.

Conclusions:

• Artificial Neural Network (ANN) is the most reliable model for potato yield forecasting in the studied region.
• Tailoring machine learning models to local agricultural conditions significantly improves prediction accuracy.
• Accurate yield predictions enable proactive planning for food security, market stability, and resource management in agriculture.