OU ACS-5513: Machine Learning Project
← Return to Introduction

Model

Model Training

Configure benchmarking parameters for clinical classification.

Deployment note: The Cleveland dataset, model artifacts, and registry are bundled with this deployment. Uploaded datasets remain runtime-only.

Model views

Choose between dataset management and model training workflows.

Cleveland Benchmark

Three models. One clinical baseline.

The bundled Cleveland cases are evaluated with the same 13-feature schema, median imputation, and fixed five-fold stratified validation. The saved estimators are fitted on every available row after validation.

Active dataset Cleveland Heart Disease Cleaned Dataset 303 rows · 13 features · target: binary diagnosis

Registered results

Validation snapshot

3 / 3 models ready
01 Top F1

Naive Bayes

83.8%

mean F1 score ± 4.2%

Accuracy
85.5% ± 4.0
Precision
86.3% ± 7.3
Recall
82.0% ± 5.1
303 rows 5 folds Ready
02

K-Nearest Neighbors

80.0%

mean F1 score ± 3.2%

Accuracy
82.2% ± 2.6
Precision
82.5% ± 2.4
Recall
77.7% ± 5.1
303 rows 5 folds Ready
03

Support Vector Machine

81.5%

mean F1 score ± 4.7%

Accuracy
83.8% ± 3.9
Precision
85.3% ± 4.9
Recall
78.4% ± 6.5
303 rows 5 folds Ready

Scores are means across stratified folds. Standard deviations are shown beside each metric; the production artifact uses median imputation inside its pipeline.

Naive Bayes K-Nearest Neighbors Support Vector Machine