---
annotations_creators:
- expert-generated
language:
- en
license:
- cc0-1.0
multicity:
- false
size_categories:
- 100K<n<1M
source_datasets:
- original
task_categories:
- tabular-classification
- tabular-regression
task_ids:
- synthetic-data
pretty_name: "SynData medical research Dataset (500000 rows)"
dataset_info:
  features:
    - name: patient_guid
      dtype: string
      description: "Patient GUID"
    - name: age_years
      dtype: float64
      description: "Age (Years)"
    - name: gender_cat
      dtype: string
      description: "Gender Category"
    - name: bmi_index
      dtype: float64
      description: "BMI Index"
    - name: systolic_bp_mmHg
      dtype: float64
      description: "Systolic BP (mmHg)"
    - name: crp_biomarker_mg_L
      dtype: float64
      description: "C-Reactive Protein (mg/L)"
    - name: hba1c_pct
      dtype: float64
      description: "HbA1c (%)"
    - name: treatment_arm
      dtype: string
      description: "Treatment Arm"
    - name: survival_months
      dtype: float64
      description: "Follow-up Survival (Mo)"
    - name: outcome_flag
      dtype: bool
      description: "Primary Endpoint Met"
    - name: time_series_vitals
      dtype: string
      description: "Vitals Telemetry Series"
  splits:
  - name: train
    num_bytes: 130799834
    num_examples: 500000
tags:
- synthetic
- differential-privacy
- monte-carlo
- gemini-ai
- tabular
- huggingface
---

# Dataset Card for SynData medical research Synthetic Dataset

## Dataset Description

- **Homepage:** https://huggingface.co/datasets/my-org-or-user/syndata-medical-research
- **Repository:** my-org-or-user/syndata-medical-research
- **Point of Contact:** SynData Studio (my-org-or-user)
- **Dataset Size:** 500000 Rows / Columns: 11
- **Generation Method:** Multivariate Monte Carlo Gaussian Copula
- **Differential Privacy Noise (Epsilon):** ε = 0.1
- **PII Leakage Status:** 0% (Clean Anonymized Synthetic Generation)

### Dataset Summary

This dataset is a fully synthetic, high-fidelity benchmark generated by **SynData Studio**. It maintains realistic mathematical covariance, non-linear feature correlations, and multi-modal time-series telemetry while guaranteeing zero exposure of real-world Personally Identifiable Information (PII) or Protected Health Information (PHI).

Designed for zero-shot ML training, synthetic pre-training, algorithm benchmarking, and cross-border privacy-preserving analytics.

## Dataset Structure

### Data Fields

- **`patient_guid`** (`uuid`): Patient GUID - De-identified HIPAA patient identifier
- **`age_years`** (`number`): Age (Years) - Participant age (18-88)
- **`gender_cat`** (`enum`): Gender Category - Biological sex assigned at birth
- **`bmi_index`** (`number`): BMI Index - Body Mass Index
- **`systolic_bp_mmHg`** (`number`): Systolic BP (mmHg) - Blood pressure reading
- **`crp_biomarker_mg_L`** (`number`): C-Reactive Protein (mg/L) - Inflammation marker
- **`hba1c_pct`** (`number`): HbA1c (%) - Glycated hemoglobin
- **`treatment_arm`** (`enum`): Treatment Arm - Trial cohort assignment
- **`survival_months`** (`number`): Follow-up Survival (Mo) - Observed trial period
- **`outcome_flag`** (`boolean`): Primary Endpoint Met - Boolean treatment response indicator
- **`time_series_vitals`** (`timestamp_series`): Vitals Telemetry Series - Multi-modal time-stamped ECG & SpO2 readings

### Sample Python Code to Load

```python
from datasets import load_dataset
import pandas as pd

# Load synthetic dataset directly from Hugging Face Hub
ds = load_dataset("my-org-or-user/syndata-medical-research", split="train")

# Convert to Pandas DataFrame
df = ds.to_pandas()
print(df.head())
```

## Direct Python Upload Script

```python
# Install required libraries
# pip install huggingface_hub datasets pandas

import pandas as pd
from datasets import Dataset

# Load local synthetic dataset
df = pd.read_json("syndata_medical_research_500000_records.json")

# Convert to Hugging Face Dataset object
hf_dataset = Dataset.from_pandas(df)

# Push to Hugging Face Hub
hf_dataset.push_to_hub("my-org-or-user/syndata-medical-research", private=False)
```

## Licensing Information

Dedicated to the public domain under **CC0 1.0 Universal (CC0 1.0) Public Domain Dedication**. Free for commercial and non-commercial model training without restriction.