Breast Cancer Diagnosis using Machine Learning

This project explores multiple machine learning models to predict breast cancer diagnosis (Benign vs. Malignant) using the Wisconsin Breast Cancer Diagnostic dataset. We implement feature engineering, model evaluation, interpretability using LIME, and advanced hyperparameter tuning.

Dataset

• Source: UCI Machine Learning Repository
• Name: Breast Cancer Wisconsin (Diagnostic)
• Features: 30 numeric features extracted from cell nuclei
• Target: Diagnosis (M = Malignant, B = Benign)

Models Used

Model	Accuracy	Precision	Recall	F1 Score	AUC-ROC
Logistic Regression	94.74%	0.9737	0.8810	0.9250	0.9921
Random Forest	97.37%	1.0000	0.9286	0.9630	0.9929
Support Vector Machine	90.35%	1.0000	0.7381	0.8493	0.9808
K-Nearest Neighbors	91.23%	0.9706	0.7857	0.8684	0.9547
Gradient Boosting	96.49%	1.0000	0.9048	0.9500	0.9947
Naive Bayes	93.86%	1.0000	0.8333	0.9091	0.9934

Key Features

• Data Preprocessing
  Missing value check, label encoding, feature scaling

• Model Building
  Logistic Regression, Random Forest, SVM, KNN, Naive Bayes, Gradient Boosting

• Model Interpretation (Explainability)
  Used LIME (Local Interpretable Model-Agnostic Explanations) for individual predictions

• Model Evaluation
  Accuracy, Precision, Recall, F1 Score, ROC-AUC, comparison bar plots

• Hyperparameter Tuning
  Used GridSearchCV for Random Forest and Gradient Boosting

• (Optional) Fairness Check
  You can extend to check model bias against demographic groups (if features are available)

Visualization

• LIME explanations of local predictions
• Bar charts comparing evaluation metrics across models
• Confusion matrices

Installation

pip install -r requirements.txt