A field-based recommender system for crop disease detection using machine learning

Area of Science:

• Agricultural Science
• Computer Science
• Information Technology

Background:

• Smallholder farmers require accessible tools for timely crop disease diagnosis and management.
• Effective agricultural practices and information are vital for agricultural sector growth and food security.

Purpose of the Study:

• To develop and evaluate a field-based recommendation system for real-time crop disease diagnosis and advisory services for smallholder farmers.
• To leverage machine learning and natural language processing for accurate disease identification and actionable recommendations.

Main Methods:

• A question-answer based recommender system was developed using machine learning and natural language processing.
• The Sentence-BERT (RetBERT) model was employed, achieving a BLEU score of 50.8%.
• The system integrates both online and offline functionalities to accommodate limited internet access in remote farming communities.

Main Results:

• The RetBERT model demonstrated the best performance among the tested state-of-the-art algorithms.
• The system provides real-time feedback on crop disease diagnosis and tailored advisory recommendations.
• The system's performance was noted to be potentially limited by the available dataset size.

Conclusions:

• The developed field-based system shows promise for improving crop disease management among smallholder farmers.
• Successful implementation can enhance agricultural productivity and contribute to alleviating food insecurity in sub-Saharan Africa.
• Further large-scale trials are recommended to validate the system's broader applicability and impact.