Optimized Breast Cancer Classification Using PCA-LASSO Feature Selection and Ensemble Learning Strategies With Optuna Optimization

This project presents an optimized breast cancer classification system integrating dimensionality reduction, feature selection, and advanced ensemble learning models. The existing framework employs Principal Component Analysis (PCA) and LASSO for feature refinement, followed by traditional classifiers like Random Forest, SVM, Gradient Boosting, and Logistic Regression, all tuned via GridSearchCV, RandomizedSearchCV, and Optuna optimization with 3-fold cross-validation. To enhance performance, the proposed system introduces Artificial Neural Networks (ANN), XGBoost, and Decision Tree classifiers for more robust prediction. The model exhibiting the best accuracy is selected for deployment. A user-friendly web interface is developed using Flask, HTML, CSS, and JavaScript to allow real-time prediction of breast cancer status—malignant or benign—by uploading test inputs. This approach combines powerful machine learning algorithms with modern web technologies to provide accurate, fast, and accessible breast cancer diagnosis support. The final system aims to aid clinical decision-making and early detection through automation and intelligent pattern recognition.

Keywords:
Breast Cancer, PCA, LASSO, Ensemble Learning, XGBoost, ANN, Decision Tree, Optuna, Classification, Flask Web App, Feature Selection, GridSearchCV, Kaggle Dataset, Malignant, Benign