Optimizing XGBoost via mSMA_plus: A Novel Meta-Heuristic Approach for High-Precision Multiclass Dry Bean...

Area of Science:

• Agricultural Science
• Computer Science
• Data Science

Background:

• Accurate dry bean classification is vital for agriculture, food security, and seed quality.
• Traditional methods are error-prone, necessitating advanced machine learning solutions.
• Hyperparameter optimization is key to high-performance machine learning models.

Purpose of the Study:

• To propose a novel framework for optimizing Extreme Gradient Boosting (XGBoost) hyperparameters.
• To classify seven dry bean varieties using the Dry Bean Dataset.
• To employ meta-heuristic algorithms for effective XGBoost model tuning.

Main Methods:

• Systematic tuning of XGBoost parameters (learning rate, tree depth, subsampling rates).
• Utilized meta-heuristic algorithms: Slime Mould, Modified SMA (mSMA), mSMA_plus, Particle Swarm Optimization, and Grey Wolf Optimizer.
• Comparative evaluation against GridSearch and RandomSearch techniques.

Main Results:

• The mSMA_plus algorithm achieved a peak classification accuracy of 99.39% and an F1-score of 0.9939.
• The proposed framework demonstrated a significant advancement over baseline methods.
• Achieved approximately 1.15% higher accuracy than GridSearch within 507.55 seconds.

Conclusions:

• Meta-heuristic optimization significantly enhances XGBoost model performance for dry bean classification.
• The mSMA_plus algorithm offers a superior approach for hyperparameter tuning in this domain.
• This advancement contributes to improved agricultural sustainability and data-driven decision-making.