A comprehensive review of machine learning techniques on diabetes detection

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine
• Computational Biology

Background:

• Diabetes mellitus presents a growing public health concern with increasing morbidity and a decreasing average age of diagnosis.
• Traditional diagnostic methods are prone to human error, necessitating advanced detection techniques.
• Artificial intelligence (AI) offers promising solutions for improved accuracy and efficiency in disease detection.

Purpose of the Study:

• To review and discuss the application of AI, data mining, and machine learning techniques for diabetes mellitus detection.
• To compare the efficacy of various machine learning and deep learning models in diagnosing diabetes.
• To identify challenges in data inadequacy and model deployment, and explore future research directions.

Main Methods:

• Exploration of data mining algorithms like density-based spatial clustering for pattern identification.
• Utilizing machine vision for feature extraction from facial images and iris patterns (iridocyclitis).
• Comparative analysis of machine learning classifiers (SVM, logistic regression, decision trees) and deep learning models (ANN, RNN, LSTM, CNN).

Main Results:

• Various AI techniques, including machine vision and pattern recognition in iris images, are employed for diabetes detection.
• Machine learning and deep learning models demonstrate potential in improving diagnostic accuracy compared to traditional methods.
• Performance evaluation commonly involves metrics such as root-mean-square error, mean absolute error, and area under the curve.

Conclusions:

• AI-driven methods present a significant advancement in the early detection and management of diabetes mellitus.
• Challenges related to data scarcity and practical model implementation require further investigation.
• Future research focusing on novel AI approaches is expected to enhance diagnostic performance and provide deeper insights into diabetes prevalence factors.