Finding a Husband: Using Explainable AI to Define Male Mosquito Flight Differences.

Area of Science:

• Entomology
• Machine Learning
• Bioacoustics

Background:

• Mosquito-borne diseases cause significant annual mortality, necessitating novel control methods due to insecticide resistance.
• Previous research utilized near-infrared tracking for mosquito behavior analysis, leading to innovative bed net designs.
• Understanding mosquito flight dynamics is crucial for developing effective disease mitigation strategies.

Purpose of the Study:

• To develop and apply a novel machine learning methodology for analyzing mosquito flight trajectories.
• To distinguish between male, female, and paired mosquito tracks using anomaly detection and trajectory analysis.
• To gain insights into mosquito behavior, particularly during mating, for potential application in genetic control interventions.

Main Methods:

• Utilized trajectory analysis and machine learning, specifically anomaly detection, on 3D mosquito flight tracks.
• Implemented novel feature engineering techniques, segmenting tracks to focus on detailed flight behaviors.
• Employed SHAP values for model interpretation to identify key flight features differentiating sexes.

Main Results:

• Achieved a balanced accuracy of 64.5% and an ROC AUC score of 68.4% in distinguishing mosquito tracks.
• Identified specific flight features contributing to sex-based behavioral differences, validated by expert opinion.
• Demonstrated the potential of the methodology to classify different mosquito classes (sex, strain, species).

Conclusions:

• The developed methodology offers a powerful tool for analyzing insect flight patterns and behaviors.
• This approach can significantly support the development of targeted genetic mosquito control strategies, particularly those focusing on mating behaviors.
• The system's adaptability extends to various trajectory analysis domains beyond entomology.