Diabetes Prediction Using Extreme Learning Machine – An Application in Health System

Accurate and early prediction of diabetes is critical for effective healthcare management and prevention of long-term complications. Traditional diagnostic approaches rely on manual clinical analysis and laboratory testing, which are often time-consuming and may delay early intervention. Although machine learning techniques have been applied to diabetes prediction, many existing models suffer from limitations such as slow training speed, high computational complexity, sensitivity to noise, and reduced performance on high-dimensional medical datasets. To address these challenges, this paper presents an Extreme Learning Machine (ELM)–based diabetes prediction framework integrated with software engineering principles for reliable health informatics applications. The proposed system utilizes patient health parameters including glucose level, insulin, body mass index (BMI), blood pressure, age, skin thickness, and diabetes pedigree function. Data preprocessing techniques such as normalization, missing value handling, and Principal Component Analysis (PCA) are applied to enhance data quality and reduce dimensionality. The ELM model enables fast and efficient classification of patients into diabetic, non-diabetic, and borderline (pre-diabetic) categories. Experimental evaluation on hospital and Indian diabetes datasets demonstrates improved prediction accuracy and significantly reduced training time compared to traditional machine learning approaches. The system provides a scalable, efficient, and reliable solution for early diabetes detection, supporting timely medical decision-making in modern healthcare environments.