Jadwal Sholat

Memuat jadwal sholatโ€ฆ

Ilmu Komputer & AI editorial

Open AccessOA2023

Prediction of Diabetes Disease Using an Ensemble of Machine Learning Multi-Classifier Models

This study proposes a pipeline-based multi-classification framework to predict diabetes (diabetic, non-diabetic, prediabetes) using an imbalanced Iraqi dataset. The weighted ensemble model achieves high accuracy (0.9887), precision (0.9861), recall (0.9792), -score (0.9851), and AUC (0.999).
Karlo Abnoosian; R. Farnoosh; M. Behzadiยท BMC Bioinformaticsยท 2023ยท DOI 10.1186/s12859-023-05465-z

The core problem

Diabetes is a life-threatening chronic disease with increasing global prevalence. Early diagnosis and treatment are crucial to prevent severe complications. Machine learning (ML) has shown promise for diabetes diagnosis, but challenges such as limited labeled data, missing values, and dataset imbalance impede the development of accurate prediction models. The authors propose a novel pipeline-based multi-classification framework to address these challenges and improve prediction performance. The study uses the imbalanced Iraqi Patient Dataset of Diabetes to classify patients into three classes: diabetic, non-diabetic, and prediabetes.

Innovation

The proposed framework incorporates multiple pre-processing steps: duplicate sample removal, attribute conversion, missing value imputation, data normalization and standardization, feature selection, and k-fold cross-validation. Several machine learning models are implemented: k-Nearest Neighbors (k-NN), Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), AdaBoost, and Gaussian Naive Bayes (GNB). To address dataset imbalance, a weighted ensemble approach based on the Area Under the Receiver Operating Characteristic Curve (AUC) is introduced. Hyperparameter tuning is performed using grid search and Bayesian optimization. The pipeline can be represented as:

The weighted ensemble combines predictions from base models using weights proportional to their AUC scores. The ensemble output can be expressed as:

where is the weight for model based on its AUC, and is the predicted probability for class .

Introduction
Diabetes is a life-threatening chronic disease with increasing global prevalence. Early diagnosis and treatment are crucial to prevent severe complications. Machine learning (ML) has shown promise for diabetes diagnosis, but challenges such as limited labeled data, missing values, and dataset imbalance impede the development of accurate prediction models. The authors propose a novel pipeline-based multi-classification framework to address these challenges and improve prediction performance. The study uses the imbalanced Iraqi Patient Dataset of Diabetes to classify patients into three classes: diabetic, non-diabetic, and prediabetes.

Why it matters

The pipeline-based multi-classification framework successfully addresses challenges associated with limited labeled data, missing values, and dataset imbalance, leading to improved prediction performance. The use of a weighted ensemble based on AUC effectively mitigates the impact of class imbalance. The high AUC value (0.999) indicates excellent discriminative ability. The study highlights the potential of machine learning techniques in diabetes diagnosis and management. The proposed framework can serve as a valuable tool for accurate prediction and improved patient care. Further research can refine and optimize the framework and explore its applicability in diverse datasets and populations.

Who should read this

Pembaca Research Digest dan praktisi rumpun ini.

Opening member contentโ€ฆ