AN ADVANCED OPTIMIZATION STRATEGY FOR ENHANCED DIABETES MELLITUS PREDICTION UTILIZING DATA–DRIVEN ADAPTIVE ENSEMBLE SERIAL CASCADED DEEP LEARNING TECHNIQUES

View at publisher Summarise with VIRAT
Authors: Girish Kulkarni, Pavan Kumar Pagadala
Cited by: 0
References: 32

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Diabetes is a kind of metabolic disease caused due to insulin breakdown in the human body. Lifestyle choices, food choices, an unfavorable environment, and hormonal changes have resulted in an increase in new cases of diabetes in recent days. Diabetics need to be treated in the initial stage else it turns into a lifelong disease that may lead to failures in the internal organs like eyes, heart, kidneys and so on. Identifying the diabetics in the initial stages effectively reduces the danger among the individuals. A variety of techniques were employed by researchers to predict diabetes mellitus. Classical techniques face more issues due to the huge training period and also executing the training in large datasets takes more time and is complex. To overcome these difficulties, an effective ensemble learning strategy is developed. In this study, a new ensemble learning-based framework is proposed for the early prediction of diabetes mellitus. The input data are collected from the three standard data sources namely, Dataset 1 (Diabetes), Dataset 2 (Diabetes Dataset), and Dataset 3 (Gestational Diabetes Mellitus (GDM Data Set)). The collected data are fed to Adaptive Ensemble Serial Cascaded Deep Learning Network (AESC-DLNet)-based diabetes prediction. This is a method where multiple models like Recurrent Autoencoder with Sparse Attention (RAE-SA), Deep Belief Network (DBN), and Dense Bi-Directional Gated Recurrent Unit (D-Bi-GRU) are combined to improve the overall efficiency and robustness of predictions. By using a serial cascade, the output of one model is fed as input to the next model in the sequence. The reliability of the proposed AESC-DLNet is enhanced through Modified Random Variable-based Lyrebird Optimization (MRV-LO). Finally, the performance of the proposed model for diabetes prediction is compared using a comprehensive suite of evaluation metrics and benchmarked datasets.

Fields of Study & Subjects

Artificial Intelligence in Healthcare Data Mining and Machine Learning Applications Machine Learning in Healthcare

Publication details

Journal / source
Biomedical Engineering Applications Basis and Communications
Volume
39
Issue
02
Publication date
25 March 2026
Language
EN