Breast Cancer Classification Project

---

Overview

This project focuses on building an efficient machine learning pipeline to classify breast cancer cases as benign or malignant using dimensionality reduction techniques and supervised learning models. The project uses the Wisconsin Breast Cancer dataset and aims to balance high accuracy with computational efficiency.

---

Objectives

• Perform dimensionality reduction using both PCA (Principal Component Analysis) and LDA (Linear Discriminant Analysis).
• Train and evaluate multiple machine learning models for classification, including:
  • Logistic Regression
  • Support Vector Machine (SVM)
  • Random Forest
  • K-Nearest Neighbors (KNN)
  • Gradient Boosting
• Compare model performance based on key metrics:
  • Accuracy
  • Precision
  • Recall
  • F1-Score
  • AUC (Area Under the Curve)
• Provide insights and recommendations based on the results.

---

Key Results

• Dimensionality Reduction:
  • PCA reduced the dataset from 30 features to 6 principal components, retaining 90% of the variance.
  • LDA maximized class separability, achieving high accuracy with just 1 linear discriminant.
• Best-Performing Model:
  • Gradient Boosting achieved the highest mean accuracy (96.65%) and perfect AUC (1.00) on PCA-transformed data.
  • Logistic Regression on LDA-transformed data achieved 98% accuracy with an AUC of 1.00.
• Evaluation Metrics:
  • All models demonstrated high accuracy (>95%) and low error rates, indicating robust performance and effective separability between classes.

---

Project Workflow

1. Data Preprocessing:

   • Handled missing values and encoded the target variable.
   • Standardized the features for optimal model performance.

2. Dimensionality Reduction:

   • Applied PCA to reduce dimensionality and visualize variance contribution.
   • Implemented LDA to project data into a supervised lower-dimensional space.

3. Model Training and Evaluation:

   • Trained Logistic Regression, SVM, Random Forest, KNN, and Gradient Boosting on PCA-transformed data.
   • Used K-Fold Cross Validation for robust evaluation of model performance.
   • Evaluated Logistic Regression on LDA-transformed data for direct comparison.

4. Performance Comparison:

   • Summarized model performance in tables and visualized results using bar charts and ROC curves.

5. Insights and Recommendations:

   • Identified Gradient Boosting and LDA as the most effective approaches.
   • Highlighted potential areas for improvement through hyperparameter tuning and external validation.

---

Technologies Used

• Python
• Libraries:
  • pandas, numpy (Data processing)
  • matplotlib, seaborn (Data visualization)
  • scikit-learn (Dimensionality reduction, machine learning models)
  • xgboost (Gradient Boosting)

---