Machine Learning

Diabetes Prediction

Stacking ensemble classifier for early diabetes prediction using 6 base models with cross-validated hyperparameter tuning. Published at AIMV 2021 on IEEE Xplore.

Source Code Read Paper

PythonScikit-learnPandasNumPy

Architecture

Stacking Ensemble

6 base classifiers → meta-learner

Gaussian Naive Bayes

Probabilistic

Random Forest

Ensemble

Decision Tree

Tree-based

Support Vector Machine

Kernel

ANN (MLPClassifier)

Neural Network

Logistic Regression

Linear

predictions as features

Meta-Learner

Logistic Regression

Test accuracy: 74.46%

Dataset

PIMA Indians Diabetes

768 instances · 8 clinical features · binary classification

PregnanciesGlucoseBloodPressureSkinThicknessInsulinBMIDiabetesPedigreeFunctionAge

Methodology

Key Points

▸6 base classifiers with hyperparameter tuning via cross-validation

▸Stacking ensemble — base model predictions become meta-learner features

▸PIMA Indians Diabetes Dataset — 768 instances, 8 clinical features

▸Mean imputation for zero-value handling + feature scaling

▸Published at AIMV 2021 (IEEE Xplore) — 10 citations

Publication

"Diabetes Prediction, using Stacking Classifier"

V. Khilwani, V. Gondaliya, S. Patel, J. Hemnani, B. Gandhi, S.K. Bharti

AIMV 2021·IEEE Xplore·DOI: 10.1109/AIMV53313.2021.9670920·10 citations

Read on IEEE Xplore

Source Code

6 Models768 SamplesIEEE Published