Semi-automated identification of biological control agent using artificial intelligence.

Area of Science:

• Agricultural Entomology
• Machine Learning Applications
• Biological Pest Control

Background:

• Accurate identification of biological control agents is essential for effective integrated pest management (IPM).
• Field identification of beneficial mites, such as phytoseiids, is challenging for non-taxonomists.
• Machine learning offers potential solutions for biological organism identification.

Purpose of the Study:

• To develop a semi-automated method for precise identification of the biological control agent Neoseiulus barkeri Hughes.
• To apply the eXtreme Gradient Boosting (XGBoost) machine learning algorithm for phytoseiid mite identification.
• To facilitate the transfer of identification expertise between specialists and non-specialists.

Main Methods:

• Collected and analyzed 512 specimens of N. barkeri and related phytoseiid species.
• Extracted 22 quantitative morphological features from mite photomicrographs, including shield and setal lengths, and spermatheca dimensions.
• Employed eXtreme Gradient Boosting (XGBoost) for classification analysis.

Main Results:

• Achieved 100% accuracy in identifying phytoseiid mites using XGBoost.
• Identified seta j4 as a key feature for significant specimen discrimination.
• Demonstrated the potential for semi-automated identification of biological control agents.

Conclusions:

• The XGBoost model provides a highly accurate method for semi-automated identification of N. barkeri.
• This approach can bridge the knowledge gap for non-expert identification of biological control agents in IPM.
• The study lays groundwork for future fully automated identification systems for pest management.