DiabeRules: a transparent rule based expert system for managing diabetes

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Computational Biology

Background:

• Diabetes mellitus is a chronic metabolic disorder characterized by hyperglycemia, posing significant health risks including cardiovascular damage, nerve damage, and stroke.
• Early detection and management of diabetes are crucial to mitigate severe complications and reduce mortality rates.
• Existing systems often lack transparency, hindering patient understanding and adherence to management strategies.

Purpose of the Study:

• To develop an expert system, DiabeRules, for transparent and comprehensible diabetes management.
• To identify primary diabetes-related rules and risk factors using a refined decision tree approach.
• To enhance early diabetes detection and facilitate preventive measures for affected individuals.

Main Methods:

• Implementation of a hybrid decision tree (DT) for identifying diabetes-related rules.
• Refinement of the DT using a Sequential Hill Climbing approach with a customizable heuristic function.
• Evaluation of the DiabeRules system using a diabetes dataset from the UCI repository, comparing its performance against existing systems.

Main Results:

• The DiabeRules model successfully generated a transparent and comprehensible set of decision rules.
• The system effectively identified essential rules and critical risk factors associated with diabetes.
• Experimental results indicated the effectiveness of DiabeRules in diabetes management compared to current systems.

Conclusions:

• DiabeRules provides an effective and transparent approach to diabetes management through comprehensible rules.
• The system aids in early risk factor identification, empowering individuals to take preventive actions.
• This expert system demonstrates the potential of AI in improving diabetes care and patient outcomes.