Data Exploration
Clinical Feature Analysis
Investigating feature distributions and multivariate relationships within the clinical data.
Provenance: Cleveland Clinic Foundation (Processed)
The bundled clinical benchmark dataset for this project. It contains the cleaned Cleveland cases used for training, evaluation, and exploratory analysis.
Data Dictionary
| Field | Domain / Scale | Units | Description |
|---|---|---|---|
age |
Numeric (0-120) | years | Age of the patient. |
sex |
0=Female, 1=Male | binary | Biological sex of the patient. |
cp |
1=Typical, 2=Atypical, 3=Non-anginal, 4=Asymptomatic | coded | Type of chest pain experienced. |
trestbps |
Numeric (0-300) | mm Hg | Resting blood pressure on admission. |
chol |
Numeric (0-1000) | mg/dl | Serum cholesterol measurements. |
fbs |
>120 mg/dl (1=True, 0=False) | binary | Fasting blood sugar level. |
restecg |
0=Normal, 1=ST-T wave, 2=LV hypertrophy | coded | Resting electrocardiographic results. |
thalach |
Numeric (0-250) | bpm | Maximum heart rate achieved. |
exang |
1=Yes, 0=No | binary | Exercise induced angina. |
oldpeak |
Numeric (0-10) | mm | ST depression induced by exercise relative to rest. |
slope |
1=Upsloping, 2=Flat, 3=Downsloping | coded | The slope of the peak exercise ST segment. |
ca |
0-3 vessels | count | Number of major vessels colored by fluoroscopy. |
thal |
3=Normal, 6=Fixed, 7=Reversable | coded | A blood disorder called thalassemia. |
target |
0=<50% diameter, 1=>50% diameter | binary | Heart disease diagnosis status. |
The strongest 2D relationships
Ranked by normalized mutual information
NMI ranges from 0 to 1 and detects general dependence after binning continuous fields, including nonlinear relationships. Continuous pairs use a LOWESS trend; coded fields use boxplots or a count heatmap because their numeric codes are categories, not measurements.
oldpeak + slopeNMI 0.16 · r 0.58 · rho 0.59
age + thalachNMI 0.12 · r -0.39 · rho -0.39
cp + exangNMI 0.12 · r 0.38 · rho 0.42
sex + thalNMI 0.11 · r 0.38 · rho 0.38