Ensemble learning-based approach for automatic classification of termite mushrooms.

Area of Science:

• Mycology
• Bioinformatics
• Machine Learning

Background:

• Termite mushrooms are valuable edible fungi, but traditional identification methods are unreliable due to their short, seasonal growth.
• Accurate species identification is crucial for understanding and conserving these important resources.

Purpose of the Study:

• To develop a novel, automated method for classifying termite mushroom species.
• To leverage machine learning and gene sequencing for reliable identification.

Main Methods:

• Utilized Gradient Boosting machine learning and sequence encoding on Internal Transcribed Spacer (ITS) gene data.
• Trained the model using ITS sequences from NCBI and BOLD databases.
• Applied ensemble learning techniques for species classification.

Main Results:

• Achieved high accuracy (0.91) and an average AUCROC of 0.99 on the test dataset.
• Validated the model with field-collected ITS sequences from Vietnam, showing consistent results with NCBI BLAST.
• The machine learning model demonstrated effective species identification.

Conclusions:

• The proposed machine learning model offers a promising automated solution for termite mushroom species classification.
• This approach can enhance research on local growth patterns and inform conservation strategies for rare termite mushroom resources.