Healthcare AICleveland Heart Disease Dataset

Cardiovascular Heart Disease Risk Diagnostic Pipeline

Clinical machine learning diagnostic system achieving 88.89% generalization with zero variance gap

Cardiovascular Heart Disease Risk Diagnostic Pipeline dashboard visualization
Top Model & Metric

Bagging Classifier (Bootstrap Ensemble)

Generalization Test Accuracy: 88.89%

F1-Score: 80.00% | Train/Test Gap: 0.00%

Dataset Specs

303 clinical cases

13 physiological markers (BP, Cholesterol, ECG, Max HR)

Target: Presence / Absence of Heart Disease

Validated Outcome

88.89% Test Accuracy & 80.00% F1-Score with zero overfitting gap

Verified Notebook Results

Explore Code on GitHubDiscuss This Project

Technologies & Machine Learning Frameworks

PythonScikit-LearnEnsemble BaggingRandom ForestSVMPandasMatplotlib

01Business & Clinical Problem Context

Cardiovascular diseases represent the single largest global cause of mortality. Clinicians need reliable, non-invasive risk stratification to detect asymptomatic coronary artery stenosis before catastrophic cardiac events occur.

02Data Pipeline & Model Engineering

Conducted comprehensive clinical exploratory data analysis on 13 cardiovascular indicators (resting blood pressure, serum cholesterol, ST depression slope, thallium stress). Implemented 5-fold cross-validated hyperparameter tuning across 8 classification algorithms (Bagging, Random Forest, SVM, Gradient Boosting, XGBoost, KNN, Logistic Regression, Decision Tree).

03Algorithm Evaluation & Empirical Benchmarks

Algorithm / ModelTrain ScoreTest Accuracy / Score5-Fold CVF1-Score / MetricStatus
Bagging Classifier (Best Generalizer)88.89%88.89%84.75%80.00%best
Support Vector Machine (RBF, C=10)93.75%86.11%84.75%76.19%contender
Random Forest (200 Trees)92.36%86.11%84.68%76.19%contender
Logistic Regression86.11%80.56%84.73%66.67%baseline
K-Nearest Neighbors (k=6, Manhattan)88.89%80.56%86.82%66.67%baseline
Gradient Boosting (Overfit Check)100.0%80.56%77.07%69.57%baseline
XGBoost Classifier100.0%80.56%79.11%66.67%baseline
Model Selection & Generalization Strategy

While complex algorithms like Gradient Boosting and XGBoost achieved 100% accuracy on training data, they suffered from noticeable variance/overfitting on the unseen test set (dropping to 80.56%). In contrast, the Bagging Classifier achieved an identical 88.89% Train Accuracy and 88.89% Test Accuracy alongside an 80.00% F1-score. Selecting Bagging over overfitted boosted trees demonstrated robust production engineering judgment.

04Quantifiable Business & Operational Impact

Enables non-invasive outpatient screening with ~89% diagnostic concordance, reducing unnecessary invasive coronary angiographies while prioritizing patients with urgent ischemic risk.

05Challenges Encountered & Engineering Solutions

Identified that gradient boosted trees overfit the 303-patient sample size. Resolved by hyperparameter-tuning a bootstrap-aggregated Bagging Classifier with max_samples=1.0 and oob_score=True, which achieved perfect generalization without variance drift.

06Key Data & Diagnostic Insights

Chest pain type (typical vs asymptomatic angina) and ST depression (oldpeak) are the highest-weighted predictive biomarkers.
Maximum heart rate achieved (thalach) inversely correlates with coronary heart disease risk.
The number of major vessels colored by fluoroscopy (ca) provides sharp separation for multivessel CAD.
Available for Machine Learning & Data Science Opportunities

Interested in deploying ML solutions like this?

Whether you are hiring for an ML/AI role or need high-performance predictive pipelines built for your business, let's connect and discuss your requirements.

Contact AkashAll Projects
INX Future Inc. Strategic Workforce Performance ModelingTexas State Workforce Compensation & Salary Forecasting