APPLYING MACHINE LEARNING TECHNIQUES IN DETECTING BACTERIAL VAGINOSIS.

Area of Science:

• Microbiology
• Medical Informatics
• Computational Biology

Background:

• Changes in the body's microbial communities are linked to various diseases.
• Bacterial Vaginosis (BV) is a prevalent condition affecting women of childbearing age, with unknown causes.
• Understanding the factors influencing microbiota shifts is crucial for disease diagnosis and treatment.

Purpose of the Study:

• To identify critical diagnostic features for Bacterial Vaginosis (BV).
• To apply classification algorithms to selected features for improved diagnosis.
• To compare the efficacy of clinical versus medical features in BV classification.

Main Methods:

• Data was divided into clinical and medical feature sets.
• Feature selection techniques were employed to isolate important diagnostic indicators.
• Classification algorithms were utilized to assess diagnostic performance (accuracy, precision, recall, F-measure).
• Performance metrics and elapsed time were compared across different feature sets and algorithms.

Main Results:

• Feature selection improved or maintained classification accuracy compared to using the full dataset.
• Significant variation was observed in the number of features selected.
• Classification algorithms demonstrated superior performance when applied to the medical feature dataset.

Conclusions:

• Medical features are more effective for diagnosing Bacterial Vaginosis (BV) using classification algorithms.
• Feature selection is a valuable step in enhancing diagnostic model performance.
• Further research into the specific medical features driving BV classification is warranted.