Diabetes diagnosis using Multi-Layer Perceptron model
|
Full Text |
Pdf
|
|
Author |
Ruaa Shallal Abbas Anooz and Aws Shallal Anooz
|
|
e-ISSN |
1819-6608 |
|
On Pages
|
1473-1481
|
|
Volume No. |
20
|
|
Issue No. |
17
|
|
Issue Date |
December 15, 2025
|
|
DOI |
https://doi.org/10.59018/0925167
|
|
Keywords |
healthcare, diabetic detection, MLP, hidden neurons, PIDD dataset.
|
Abstract
Nowadays, machine learning approaches have become widely applied for diagnosing various features and are utilized across numerous fields. In healthcare, machine learning has been a key research area, where models are employed on medical databases to detect different characteristics. Detecting diabetes is particularly crucial due to the severe risks posed by the disease. While the Multi-Layer Perceptron (MLP) approach has been widely used in different applications, many recent studies lack a detailed investigation of its performance with varying network configurations, particularly the number of hidden neurons. This paper addresses this limitation by proposing a novel approach that systematically evaluates the impact of varying the number of hidden neurons in the MLP architecture. The study investigates hidden neurons ranging from 50 to 100, with an increment step of 10, to identify the optimal configuration. The samples of healthy and diabetes classes were taken from an open-sourced database named Pima Indian Diabetes Dataset (PIDD). Various performance metrics are employed to evaluate the model. Experimental results demonstrate that the proposed MLP approach achieves its highest performance with 90 hidden neurons, achieving 85.345% detection accuracy. This work provides a deeper understanding of how network configurations impact MLP performance, presenting a promising approach for non-invasive diabetes detection.
Back