Model Card for NZR – Breast Cancer Early Detection AI

This model performs binary classification to detect malignant versus benign tumors using clinical diagnostic data from the Wisconsin Breast Cancer Diagnostic Dataset. It was built with a Random Forest classifier and designed for early-stage breast cancer screening support.

This modelcard aims to be a base template for open-sourced medical AI, particularly for structured (tabular) data inputs.

Model Details

Model Description

Developed by: Alan Jafari (TekTonic AI)
Funded by [optional]: Self-funded (independent research)
Shared by: TekTonic AI
Model type: Random Forest Classifier (100 estimators)
Language(s) (NLP): Not applicable (structured data)
License: MIT
Finetuned from model: N/A (trained from scratch)

Model Sources

Repository: [Add GitHub/Kaggle link]
Paper [optional]: N/A
Demo [optional]: [Add Streamlit or web demo link if available]

Uses

Direct Use

This model can be directly used to classify breast cancer tumors (benign vs malignant) using clinical feature data. It accepts 10 numerical features as input (e.g., radius_mean, texture_mean, etc.) and returns a class prediction.

Downstream Use

Integration in medical dashboards
Research on breast cancer ML techniques
Educational applications in healthcare AI

Out-of-Scope Use

Not for use with image data (e.g., mammograms or MRI)
Not a substitute for clinical diagnosis
Not trained for male breast cancer or pediatric cases

Bias, Risks, and Limitations

Trained only on adult female data
Demographic diversity not captured
Possible overconfidence on limited data
No interpretability module (e.g., SHAP not embedded yet)

Recommendations

Use under expert medical supervision
Complementary to, not a replacement for, radiology or biopsy
Retrain with local data for domain adaptation

How to Get Started with the Model

import joblib
import numpy as np

model = joblib.load("nzr_model.pkl")

# Example input: 10 numerical diagnostic features
x_input = np.array([[14.5, 20.0, 95.0, 660.0, 0.1, 0.15, 0.08, 0.05, 0.18, 0.06]])
prediction = model.predict(x_input)

Training Details

Training Data

Dataset: Wisconsin Breast Cancer Diagnostic Dataset
Features: 10 handpicked numerical features from the full set of 30
Labels: 0 = Benign, 1 = Malignant

Training Procedure

Train/test split: 80/20
Scikit-learn RandomForestClassifier with 100 trees
Standardization and cleaning applied

Training Hyperparameters

n_estimators = 100
max_depth = None
Precision: fp32

Speeds, Sizes, Times

Training time: <10s (on CPU)
Inference latency: ~1ms
Model size: ~250 KB

Evaluation

Testing Data, Factors & Metrics

Testing Data

20% hold-out from WBCD

Factors

No demographic breakdown available

Metrics

Accuracy: 96.5%
Precision: 98%
Recall: 93%
F1-Score: 95.5%

Results

The model demonstrated high performance on classification accuracy and precision, while maintaining balance between false positives and false negatives.

Summary

Reliable early-stage diagnostic support tool for breast cancer screening based on structured input.

Model Examination

Feature importance rankings are available upon request. No black-box dependencies.

Environmental Impact

Hardware Type: CPU (Intel i7)
Hours used: ~0.01
Cloud Provider: N/A (local)
Compute Region: Iran
Carbon Emitted: Negligible

Technical Specifications

Model Architecture and Objective

Random Forest binary classifier, built using 100 decision trees. Optimized for high recall on malignant class.

Compute Infrastructure

Hardware

CPU: Intel Core i7 (local machine)

Software

Python 3.9
Scikit-learn 1.4+

Citation

BibTeX:

@misc{nzr2025,
  title={NZR: Breast Cancer Early Detection Model},
  author={Alan Jafari},
  year={2025},
  howpublished={\url{https://huggingface.co/USERNAME/nzr-model}},
}

APA: Jafari, A. (2025). NZR: Breast Cancer Early Detection Model [Machine learning model]. Hugging Face. https://huggingface.co/USERNAME/nzr-model

Glossary

Malignant: Cancerous, potentially dangerous
Benign: Non-cancerous tumor
Precision: TP / (TP + FP)
Recall: TP / (TP + FN)

More Information

Contact via t.me/alan_jafari

Model Card Authors

Alan Jafari (TekTonic AI)

Model Card Contact

Email: [email protected]