Artificial Flora Algorithm-Based Feature Selection with Gradient Boosted Tree Model for Diabetes Classification.

Area of Science:

• Medical informatics
• Machine learning
• Computational biology

Background:

• Accurate classification of diabetes mellitus is crucial for effective treatment strategies.
• Existing methods may have limitations in handling complex patient data for precise diagnosis.

Purpose of the Study:

• To develop and validate a novel model for classifying the three main types of diabetes: type I, type II, and gestational diabetes mellitus.
• To utilize patient attributes for improved diagnostic accuracy in diabetes classification.

Main Methods:

• A hybrid approach combining the Artificial Flora Algorithm (AFA) for feature selection and Gradient Boosted Trees (GBT) for classification.
• Data preprocessing included format conversion and transformation, followed by AFA-driven selection of patient features (demographics, vitals, labs, medications).
• The GBT model was trained and tested on three distinct diabetes datasets.

Main Results:

• The AFA-GBT model demonstrated high performance across three datasets.
• Achieved a maximum average precision of 91.64%, recall of 97.46%, accuracy of 99.93%, F-score of 94.19%, and kappa of 96.61%.

Conclusions:

• The developed AFA-GBT model efficiently classifies patient diagnoses into type I, type II, or gestational diabetes mellitus.
• This model offers a robust and accurate tool for medical data classification in diabetes management.