Machine Learning-Based Morphological Classification and Diversity Analysis of Ornamental Pumpkin Seeds

Area of Science:

• Agricultural Science
• Computer Science
• Data Science

Background:

• Ornamental pumpkin (Cucurbita pepo L. var. ovifera) seeds exhibit significant morphological variability, complicating classification.
• Accurate seed classification is crucial for agricultural production, breeding, and commercial seed sorting.
• Existing classification methods may lack the efficiency and accuracy required for the seed industry.

Purpose of the Study:

• To employ machine learning models for classifying ornamental pumpkin seeds based on physical and color attributes.
• To evaluate the performance of Random Forest (RF), LightGBM, and k-Nearest Neighbors (KNN) algorithms.
• To determine the most effective model for automated seed classification in the ornamental pumpkin industry.

Main Methods:

• Collected morphological (mass, elongation, width, thickness) and colorimetric (CIELAB L*, a*, b*) data from 900 ornamental pumpkin seeds (six varieties).
• Preprocessed the dataset using normalization and balancing techniques to optimize model training.
• Trained and evaluated RF, LightGBM, and KNN models, assessing performance with metrics like Accuracy, Precision, Recall, F1 Score, MCC, and Cohen's Kappa.

Main Results:

• The Random Forest (RF) model demonstrated superior performance, achieving 95.9% accuracy and a Matthews Correlation Coefficient (MCC) of 0.951.
• RF model achieved high scores across all metrics: Balanced Accuracy (0.961), Precision (0.962), Recall (0.961), and F1 Score (0.961).
• The k-Nearest Neighbors (KNN) model exhibited the lowest classification performance among the tested algorithms.

Conclusions:

• Machine learning, particularly the Random Forest model, offers a highly accurate and efficient solution for automated ornamental pumpkin seed classification.
• AI-driven classification can significantly minimize errors, enhance efficiency in seed sorting, and support the development of effective breeding schemes.
• This approach holds substantial potential for advancing the ornamental pumpkin seed industry through improved quality control and selection processes.