Java-based diabetes type 2 prediction tool for better diagnosis.

Area of Science:

• Computational intelligence
• Machine learning in healthcare

Background:

• Diabetes diagnosis is a critical challenge, often hindered by the complexity of clinical data interpretation.
• Existing neural network classification methods show promise but lack patient accessibility.
• There is a need for user-friendly tools to facilitate early diabetes detection.

Purpose of the Study:

• To develop an accurate and accessible diagnostic tool for type 2 diabetes.
• To leverage computational intelligence for improved diabetes prediction.
• To create a system that assists medical practitioners in patient diagnosis.

Main Methods:

• A neural network mixture of experts model was employed.
• The model was trained using the expectation-maximization algorithm on a dataset of 768 individuals from the Pima Indian population.
• A graphical user interface was developed for intuitive data input and output.

Main Results:

• The mixture of experts model achieved a diagnostic accuracy of 97%.
• The developed graphical user interface effectively presents the prediction results.
• The system provides instant diabetes predictions with a high degree of confidence.

Conclusions:

• The study presents a machine-implementable approach for minimizing diagnostic errors in diabetes.
• The developed tool empowers physicians to quickly diagnose type 2 diabetes by entering patient health profiles.
• Further research replicating these findings in diverse populations could enhance diagnostic capabilities globally.