Machine learning in classification and identification of nonconventional vegetables

Area of Science:

• Agricultural Science
• Computer Science

Background:

• Nonconventional vegetables offer unique nutritional and economic value, yet their identification and utilization require further research.
• Reintroducing these species enhances their appeal due to nutritional quality and nostalgic connections.
• Encouraging production and consumption of diverse vegetables is crucial globally and in Brazil.

Purpose of the Study:

• To apply machine learning for supervised classification and identification of five nonconventional leafy vegetables.
• To assess the efficacy of machine learning models using macro and micronutrient profiles.
• To validate classifier performance using cross-validation via Monte Carlo simulation.

Main Methods:

• Supervised classification models were developed using machine learning techniques.
• Nutritional characteristics (macro and micro nutrients) were used as features for classification.
• Monte Carlo simulation-based cross-validation was employed to evaluate model performance over ten replications.

Main Results:

• Machine learning models demonstrated a high success rate, exceeding 89% for most classifiers tested.
• The study successfully classified and identified nonconventional vegetables using limited nutritional attributes.
• Key performance metrics including false positive/negative rates, sensitivity, and accuracy were analyzed.

Conclusions:

• Machine learning is a viable technique for the classification and identification of nonconventional vegetables.
• Accurate identification can be achieved with a minimal set of nutritional attributes.
• This approach supports the increased production and consumption of valuable nonconventional vegetable species.