A supportive attribute-assisted discretization model for medical classification.

Area of Science:

• Medical Informatics
• Data Science
• Machine Learning

Background:

• Discretization of continuous-valued symptoms is essential for medical classification.
• Existing methods may lose information or increase data uncertainty.

Purpose of the Study:

• To propose a novel discretization model, supportive attribute-assisted discretization (SAAD), for medical diagnostic problems.
• To enhance medical classification accuracy by minimizing information loss and data uncertainty.

Main Methods:

• SAAD identifies the most supportive symptom correlating with the continuous-valued symptom.
• Discretization is performed using this supportive information and class attribute interaction.
• The model intelligently handles each continuous-valued symptom differently.

Main Results:

• Empirical experiments on ten real-life UCI datasets were conducted.
• SAAD was compared against state-of-the-art discretization approaches using various classifiers.
• The proposed approach demonstrated significant improvements in classification accuracy.

Conclusions:

• SAAD effectively enhances diagnostic accuracy in medical classification tasks.
• The model's intelligent handling of symptom interactions minimizes information loss and uncertainty.
• SAAD proves to be a valuable tool for medical data preprocessing.