Machine Learning and Data Mining Methods in Diabetes Research.

Area of Science:

• Biotechnology and Health Sciences
• Computational Biology
• Medical Informatics

Background:

• Advances in biotechnology generate vast amounts of health data, including genetic and Electronic Health Records (EHRs).
• Diabetes mellitus (DM) is a global health concern, producing extensive research data.
• Machine learning (ML) and data mining are essential for extracting knowledge from this data.

Purpose of the Study:

• To systematically review ML and data mining applications in diabetes research.
• To categorize applications into prediction/diagnosis, complications, genetics, and healthcare management.
• To identify popular methods and data types used in diabetes research.

Main Methods:

• Systematic literature review of ML and data mining in diabetes research.
• Analysis of algorithm types (supervised vs. unsupervised) and specific algorithms used.
• Examination of data sources, primarily clinical datasets from EHRs.

Main Results:

• Prediction and Diagnosis was the most popular application area.
• Supervised learning approaches dominated (85%), with unsupervised methods (15%) including association rules.
• Support Vector Machines (SVM) emerged as the most successful and frequently used algorithm.
• Clinical datasets were predominantly utilized.

Conclusions:

• ML and data mining are vital for transforming diabetes data into actionable knowledge.
• The reviewed applications demonstrate the utility of these methods for hypothesis generation and deeper understanding of DM.
• Further investigation into ML/data mining applications can enhance diabetes research and patient care.