Breast Cancer Prediction Using Machine Learning

Project Overview

Breast Cancer Prediction Using Machine Learning is a Python-based machine learning project designed to classify breast tumors as benign or malignant. The application uses the Breast Cancer Wisconsin Diagnostic Dataset and analyzes 30 medical features related to tumor characteristics.

The project demonstrates the complete machine learning workflow, including data preprocessing, feature transformation, model training, model evaluation, best-model selection and real-time prediction through a Flask web application.

Available Features

• Breast tumor classification
• Benign and malignant prediction
• 30 medical input features
• Breast Cancer Wisconsin Diagnostic Dataset
• Data preprocessing
• Missing-value handling
• Categorical feature encoding
• Feature scaling
• Feature engineering
• Multiple machine learning model training
• Model performance comparison
• Accuracy evaluation
• Precision evaluation
• Recall evaluation
• F1-score evaluation
• Best model selection
• Real-time prediction
• Flask web interface
• Prediction pipeline
• Training pipeline
• Logging
• Exception handling
• Saved trained model

Dataset

The project uses the Breast Cancer Wisconsin Diagnostic Dataset available through the scikit-learn ecosystem. The classification problem contains 30 medical features and two prediction classes: benign and malignant.

The features include measurements related to tumor radius, texture, perimeter, area, smoothness, compactness, symmetry, fractal dimension and corresponding standard-error and worst-value measurements.

Machine Learning Models

Multiple classification algorithms are trained and evaluated to identify the best-performing model for deployment.

• Logistic Regression
• Support Vector Machine
• Gaussian Naive Bayes
• Random Forest Regressor
• Gradient Boosting
• Decision Tree
• Multi-Layer Perceptron Neural Network

The project evaluates the trained models using accuracy, precision, recall and F1-score before selecting the best-performing model for final predictions.

Data Preprocessing

The data preprocessing pipeline loads and cleans the dataset, handles missing values, encodes categorical information, scales numerical features and performs the required feature transformations before model training.

Training Pipeline

The training pipeline prepares the dataset, performs preprocessing, trains multiple classification models, evaluates their performance and stores the selected best-performing model for later prediction.

Prediction Pipeline

The prediction pipeline loads the trained model and applies the same required transformations to user-provided medical parameters. The processed input is then passed to the selected model to generate a benign or malignant prediction.

Flask Web Application

The project includes a Flask-based web interface where users can enter the required medical parameters. The application processes the submitted values through the prediction pipeline and displays the resulting tumor classification.

Model Evaluation

Model performance is evaluated using multiple classification metrics. Accuracy measures overall correct predictions, while precision, recall and F1-score provide additional insight into classification performance.

Project Components

• Data Ingestion: Loads and prepares the dataset.
• Data Transformation: Performs preprocessing and feature engineering.
• Model Trainer: Trains and evaluates machine learning models.
• Training Pipeline: Coordinates the model training workflow.
• Prediction Pipeline: Handles real-time prediction.
• Logger: Records application and pipeline activities.
• Exception Handler: Provides structured error handling.
• Utilities: Contains reusable project functions.

Model Artifact Management

The trained model and processed datasets are stored as project artifacts. The selected model is saved in pickle format so that it can be loaded by the prediction pipeline without retraining the complete system for every prediction.

Notebook and Data Analysis

The project includes Jupyter Notebook resources for data exploration, visualization and model training experiments. These notebooks help understand the dataset and evaluate different machine learning approaches before deployment.

Logging and Error Handling

The application includes logging functionality for tracking project execution and model pipeline activities. A custom exception component is also included to make errors easier to identify and debug.

Web Prediction Workflow

1. Open the Flask prediction interface.
2. Enter the required medical parameters.
3. Submit the prediction form.
4. The input data is processed through the prediction pipeline.
5. The trained machine learning model analyzes the processed features.
6. The application returns the predicted tumor classification.

Future Enhancements

• Deploy the prediction model as a cloud-based API.
• Train the model using a larger and more diverse dataset.
• Explore deep learning techniques for advanced feature extraction.
• Extend the application with additional healthcare prediction capabilities.

Technology Stack

• Python
• Flask
• Scikit-learn
• NumPy
• Pandas
• Jupyter Notebook
• HTML
• CSS
• Pickle

Software and Tools Required

• Python
• Visual Studio Code
• Jupyter Notebook
• Web Browser
• pip

Important Medical Disclaimer

This project is an educational machine learning application based on the Breast Cancer Wisconsin Diagnostic Dataset. Its predictions should not be treated as a medical diagnosis or a replacement for professional medical evaluation.

Watch Project Demo

Watch Breast Cancer Prediction Using Machine Learning Demo

More Details and Full Documentation

View Breast Cancer Prediction Using Machine Learning Details