Incorporating Machine Learning Techniques to Enhance Rodent Surveillance in Marginalized Urban Communities.

Area of Science:

• Ecology
• Computer Science
• Public Health

Background:

• Rodent pest management requires efficient population surveillance.
• Current methods like trapping and manual tracking plate analysis are costly and time-consuming.
• Interpreting tracking plates demands significant expertise and time.

Purpose of the Study:

• To develop and evaluate Machine Learning (ML) techniques for analyzing rodent tracking plates.
• To compare the accuracy of ML-based analysis with conventional human interpretation.
• To provide a more efficient and cost-effective rodent surveillance method.

Main Methods:

• Image processing techniques (Otsu method, global thresholding) were used to prepare tracking plate images.
• Dimensionality reduction methods (Principal Component Analysis - PCA, Independent Component Analysis - ICA, Legendre Moments - LM) were applied.
• K-nearest neighbors (k-NN) classification was used to predict feature vectors from PCA, ICA, and LM results.

Main Results:

• PCA and LM methods showed favorable comparison against the conventional manual interpretation.
• The ML approach offers a timely and cost-effective alternative for rodent surveillance.
• Identified key patterns on tracking plates for effective analysis.

Conclusions:

• ML integration significantly enhances rodent surveillance protocols.
• This novel approach is particularly beneficial for Low- and Middle-Income Countries.
• Improved surveillance aids in managing rodent distribution hotspots and controlling rodent-borne zoonoses.