A random forest classifier for lymph diseases

Area of Science:

• Medical Informatics
• Computational Biology
• Machine Learning in Healthcare

Background:

• Machine learning aids healthcare decisions in diagnosis and prognosis.
• High dimensionality in datasets hinders classification performance.
• Feature selection is crucial for improving accuracy with limited data.

Purpose of the Study:

• To develop a robust system for diagnosing lymph diseases using machine learning.
• To investigate the effectiveness of various feature selection methods for high-dimensional medical data.
• To optimize classification accuracy by reducing feature space dimensionality.

Main Methods:

• Proposed a two-stage system: feature selection followed by classification.
• Employed diverse feature selection algorithms: genetic algorithm (GA), PCA, Relief-F, Fisher, SFFS, SBFS.
• Utilized a random forest classifier (RFC) for disease classification on reduced datasets.

Main Results:

• The genetic algorithm (GA) effectively reduced the feature space from eighteen to six dimensions.
• The GA-RFC model achieved the highest classification accuracy at 92.2%.
• Feature selection significantly improved the efficiency and performance of lymph disease diagnosis.

Conclusions:

• The GA-RFC approach demonstrates high efficacy in diagnosing lymph diseases.
• Feature selection is a critical step for enhancing machine learning model performance in healthcare.
• This method offers a promising tool for clinical decision support in lymph disease diagnosis.