All work
Academic Research · Classification
Heart Disease Classification
Compared logistic regression, K-nearest neighbors and decision trees for heart-disease diagnosis using predictive and interpretability criteria.
Why this matters
Shows structured model comparison and the ability to explain why the statistically strongest model is not always the most interpretable one.
The question
Problem
Evaluate which patient characteristics predict heart-disease diagnosis and compare multiple classification approaches.
The work
Approach
- 01
Prepared a dataset of 918 observations and 11 features.
- 02
Evaluated goodness of fit, cutoff selection, confusion matrices and ROC curves.
- 03
Compared model accuracy with transparency and implementation trade-offs.
Verified evidence
What the analysis surfaced
Logistic AUC
Decision-tree accuracy
KNN accuracy
The decision
Recommended action
- Use logistic regression when overall discriminatory performance is the priority.
- Use a decision tree when transparent decision rules are more important.
Capabilities demonstrated
ClassificationROC analysisModel interpretationClinical-data EDA
Tools
RLogistic regressionKNNDecision treeROC analysis
Limits & responsible use
- This academic analysis is not a clinical diagnostic tool.