Coupled online sequential extreme learning machine model with ant colony optimization algorithm for wheat yield...

Area of Science:

• Agricultural Science
• Machine Learning
• Data Science

Background:

• Inaccurate agricultural planning and yield predictions lead to inflated market prices and increased wheat imports.
• Effective wheat yield prediction is crucial for optimizing agricultural strategies and ensuring food security.

Purpose of the Study:

• To design and evaluate a two-phase universal machine learning model for predicting wheat yield (W_pred).
• To enhance agricultural planning and decision-making through accurate crop yield forecasting.

Main Methods:

• Developed a hybrid Ant Colony Optimization-Online Sequential Extreme Learning Machine (ACO-OSELM) model.
• Utilized a two-phase approach: ACO for feature selection and OSELM for yield prediction incorporating lagged data.
• Benchmarked ACO-OSELM against hybrid ACO-ELM and ACO-RF models using wheat yield data from Punjab province, Pakistan.

Main Results:

• The hybrid ACO-OSELM model demonstrated superior performance compared to ACO-ELM and ACO-RF models in wheat yield prediction.
• The model successfully predicted future wheat yields at test stations using historical data.
• Partial autocorrelation function was used to identify statistically significant lagged yield data for model input.

Conclusions:

• The developed two-phase ACO-OSELM model is effective for predicting wheat yield.
• This model shows significant potential as a decision-making system for crop yield prediction in data-rich agricultural regions.
• Accurate wheat yield prediction can mitigate issues related to market rates and import dependency.