Drug Recommendation System for Diabetes Using a Collaborative Filtering and Clustering Approach: Development and...

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Data Mining for Public Health

Background:

• Diabetes mellitus is a significant global public health concern requiring continuous management.
• Effective diabetes control involves personalized treatment strategies and accurate medication selection.
• Healthcare institutions possess vast patient data amenable to analysis for improved treatment outcomes.

Purpose of the Study:

• To develop a drug recommendation system for diabetes patients.
• To utilize collaborative filtering and clustering techniques as adjuncts to physician-prescribed treatments.
• To enhance diabetes management through data-driven personalized medication suggestions.

Main Methods:

• Applied data mining techniques to a diabetes patient dataset from the UCI Machine Learning Repository.
• Employed unsupervised learning for dimensionality reduction and patient clustering.
• Utilized user-based collaborative filtering to generate drug recommendations based on patient profiles and group characteristics.

Main Results:

• Principal Component Analysis identified eight components explaining data variability.
• Clustering algorithm successfully segmented patients into six distinct groups.
• The recommendation system demonstrated acceptable performance with a Mean Squared Error of 0.51 and recommendation accuracy of 0.61.

Conclusions:

• The developed system effectively suggests medications for diabetes patients based on their characteristics and drug information.
• Clustering and prediction techniques provide acceptable results for the recommendation process.
• This technology offers a novel approach to support healthcare professionals in managing diabetes treatment and control.