Identification of medically and forensically relevant flies using a decision treelearning method

Area of Science:

• Forensic Entomology
• Machine Learning Applications
• Insect Taxonomy

Background:

• Accurate identification of flies (blow flies, flesh flies, house flies) is crucial for forensic science, public health, and animal health.
• Traditional morphological and molecular identification methods face limitations with large specimen numbers.
• Machine learning offers promising solutions for insect classification challenges.

Purpose of the Study:

• To develop and evaluate a machine learning model for discriminating between three families and seven species of flies.
• To assess the accuracy of a decision tree model using wing morphometric data for fly identification.

Main Methods:

• Applied a decision tree machine learning algorithm.
• Utilized wing morphometric data for fly classification.
• Tested the model on seven fly species across three families: Calliphoridae, Sarcophagidae, and Muscidae.

Main Results:

• Achieved 100% accuracy in identifying flies at the family level.
• Reached 83.33% accuracy in identifying flies at the species level.
• Demonstrated the potential for a non-expert identification tool.

Conclusions:

• Wing morphometric data and decision tree models show high accuracy for fly family and species identification.
• The developed tool has potential for use by non-experts in forensic and public health contexts.
• Further research with more species and samples is recommended for broader application, particularly for Thai fly species.