Research on features selection of medical diagnostic models based on L-S-ACO algorithm

Area of Science:

• Machine Learning
• Medical Informatics
• Computational Intelligence

Background:

• Feature selection in medical diagnostics requires addressing feature interdependencies.
• Ant Colony Optimization (ACO) is effective for feature selection and correlation challenges.
• Existing ACO methods show improved performance when integrated with other algorithms.

Purpose of the Study:

• To develop a novel algorithm, L-S-ACO, for feature selection in medical diagnostic models.
• To address feature correlation issues in medical data using an integrated approach.
• To evaluate the performance of a diagnostic model based on the proposed L-S-ACO algorithm.

Main Methods:

• The study introduces the L-S-ACO algorithm, combining Ant Colony Optimization (ACO) with Light Gradient Boosting Machine (LightGBM) and Stochastic Average Gradient (SAG).
• A medical diagnostic model was developed utilizing the L-S-ACO algorithm.
• The practical application and performance of the model in disease diagnosis were analyzed.

Main Results:

• The L-S-ACO algorithm demonstrated significant improvements in diagnostic model performance.
• Accuracy (ACC) increased by 5.81%, F1 score by 6.21%, and Area Under the Curve (AUC) by 4.08%.
• Precision improved by 5.12% and Recall by 6.35% compared to baseline methods.

Conclusions:

• The proposed L-S-ACO algorithm effectively handles feature interdependencies and correlations in medical data.
• The developed diagnostic model shows enhanced accuracy and reliability for disease diagnosis.
• This integrated approach offers a promising advancement in medical diagnostic modeling.