Machine Learning for Early Breast Cancer Detection: A Complete Practice from Data to Deployment

A breast cancer prediction system based on logistic regression, trained using the UCI Wisconsin dataset, providing a web interface for real-time prediction and deployed on the Render cloud platform.

• Original Author/Maintainer: Poornima Sonkar (GitHub, LinkedIn)
• Source Platform: GitHub
• Original Title: Breast-Cancer-Prediction
• Original Link: https://github.com/poornimasonkar/Breast-Cancer-Prediction
• Publication Date: May 26, 2026

---

Breast cancer is one of the most common malignant tumors among women worldwide, and early detection is crucial for improving the cure rate. Traditional diagnostic methods rely on doctors' experience and pathological analysis, while the introduction of machine learning technology provides new possibilities for auxiliary diagnosis. This project demonstrates a complete machine learning application development process, from data preprocessing to model deployment, providing a practical reference case for learners in the field of medical AI.

---

This project uses the Breast Cancer Wisconsin (Diagnostic) Dataset from the UCI Machine Learning Repository, which is one of the most classic medical datasets in the field of machine learning.

Dataset Features:

• Sample Source: Fine needle aspiration biopsy images of breast masses
• Number of Features: 30 numerical features describing the morphological characteristics of cell nuclei
• Target Classes: Malignant and Benign
• Feature Dimensions: Including radius, texture, perimeter, area, smoothness, compactness, concavity, concave points, symmetry, fractal dimension, etc.

These features are extracted from digitized images and can quantify the geometric and texture properties of cell nuclei, providing reliable input for machine learning models.

---

The project uses Logistic Regression as the classification algorithm. This choice reflects a pragmatic engineering mindset— in medical diagnosis scenarios, model interpretability is often more valuable than complex black-box models. Logistic regression can not only provide prediction results but also output probability values, helping doctors understand the confidence of the prediction.

Layer	Technology	Function
Frontend	HTML/CSS	User interaction interface
Backend	Flask	Web service framework
Model	scikit-learn	Machine learning algorithm library
Deployment	Render.com	Cloud platform hosting
Serialization	Pickle	Model saving and loading

---

The workflow of the entire prediction system is designed to be concise and clear:

1. Data Input: Users enter 30 tumor feature values in the web interface
2. Feature Transmission: The frontend sends data to the Flask backend service
3. Model Inference: The pre-trained logistic regression model performs prediction calculations
4. Result Display: The system returns the diagnosis result of "Benign" or "Malignant"

This end-to-end workflow design allows medical staff without technical backgrounds to use it easily, lowering the technical threshold for AI-assisted diagnosis.

---

The project not only includes model training code but also provides a complete web application and deployment plan. Learners can learn from it:

• Data preprocessing and feature engineering
• Model training and evaluation (accuracy, confusion matrix)
• Web application development
• Cloud platform deployment practice