tarekmasryo/cancer-risk-factors-data

🧬 Cancer Risk Factors & Types (2,000 Rows)

Author: Tarek Masryo · Kaggle
License: CC BY 4.0 (Attribution) — Free for research, education, and commercial use

---

📌 Dataset Summary

Clean, standardized tabular dataset linking lifestyle, environmental, and genetic factors to five cancer types.

• 2,000 rows × 21 columns
• Encodings: ordinal exposure indices (0–10), demographics (Age, BMI, Gender), binary flags (0/1) for family/genetics/infection
• Includes engineered fields: Overall_Risk_Score ∈ [0,1] and Risk_Level ∈ {Low, Medium, High}

Ideal for:

• EDA and interactive dashboards
• Multiclass tabular ML (Cancer_Type)
• Class-imbalance handling and reporting beyond accuracy (e.g., macro-F1)
• Teaching feature engineering and ordinal/binary encodings

---

🗂 Dataset Structure

Main file: data/cancer-risk-factors.csv (one row per individual)

Targets & Tasks

• Primary target: Cancer_Type ∈ {Lung, Breast, Colon, Prostate, Skin} — multiclass classification
• Optional task: Risk_Level derived from Overall_Risk_Score
  • Default thresholds: Low < 0.35, 0.35 ≤ Medium ≤ 0.65, High > 0.65

Data splits

• Single CSV; when loaded via 🤗 Datasets, it appears as a train split by default.

---

📑 Data Dictionary

Column	Description
Patient_ID	Unique individual identifier
Age	Age in years
Gender	0 = Female, 1 = Male
BMI	Body Mass Index
Smoking	Ordinal exposure index (0–10)
Alcohol_Use	Ordinal exposure index (0–10)
Obesity	Ordinal index (0–10) of obesity-related risk behaviors
Diet_Red_Meat	Ordinal intake frequency (0–10)
Diet_Salted_Processed	Ordinal intake frequency (0–10)
Fruit_Veg_Intake	Ordinal servings/frequency (0–10)
Physical_Activity	Ordinal behavior/frequency (0–10)
Physical_Activity_Level	Self-rated activity level (0–10)
Air_Pollution	Ordinal exposure index (0–10)
Occupational_Hazards	Ordinal exposure index (0–10)
Calcium_Intake	Ordinal intake frequency (0–10)
Family_History	0/1 — positive family history
BRCA_Mutation	0/1 — BRCA mutation flag
H_Pylori_Infection	0/1 — Helicobacter pylori infection
Overall_Risk_Score	Composite risk score in [0,1] (higher with higher exposures)
Risk_Level	Low / Medium / High (thresholds above)
Cancer_Type	Lung / Breast / Colon / Prostate / Skin

0–10 indices represent ordinal intensity/frequency (0 = none/very low … 10 = very high).

---

🚀 Usage

Load with 🤗 Datasets (recommended)

from datasets import load_dataset

ds = load_dataset("tarekmasryo/cancer-risk-factors-data")
df = ds["train"].to_pandas()

print(df.shape)
print(df.head())

Load with pandas (works anywhere)

import pandas as pd
from huggingface_hub import hf_hub_download

path = hf_hub_download(
    repo_id="tarekmasryo/cancer-risk-factors-data",
    repo_type="dataset",
    filename="data/cancer-risk-factors.csv",
)

df = pd.read_csv(path)
print(df.shape)

---

🧪 Quick Modeling Tips

• Use stratified splits/CV for Cancer_Type (multiclass) and for Risk_Level (imbalanced).
• Report macro-F1 alongside accuracy; add a per-class confusion matrix.
• For tree-based models, 0–10 can remain integers; for linear models, consider scaling.
• Consider class weights or focal loss-style approaches if you see strong imbalance.

---

🧭 Ethical Considerations

Educational/research dataset only — not for clinical use, diagnosis, or treatment.

---

📚 Citation

Tarek Masryo. (2025). Cancer Risk Factors & Types (2,000 Rows). Hugging Face Datasets.

📜 License

CC BY 4.0 (Attribution) — Free to share and adapt with attribution.