🧠 Breast Cancer Detection Using ML (Logistic Regression, KNN, SVM) This project uses machine learning algorithms to classify tumors as Malignant (M) or Benign (B) using the Breast Cancer Wisconsin dataset. The steps include data preprocessing, exploratory data analysis (EDA), feature selection, scaling, and model building with evaluation.
📂 Dataset Source: breast_cancer.csv
Total entries: 569
Diagnosis labels:
M = Malignant (encoded as 1)
B = Benign (encoded as 0)
🔍 Workflow Overview
- Data Preprocessing Removed unnecessary id column.
Encoded target variable: M → 1, B → 0.
Checked for missing values using missingno.
Normalized features using StandardScaler.
- Exploratory Data Analysis Visualized class distribution.
Plotted distribution of all features.
Used a heatmap to observe correlation among features.
- Feature Selection Removed features having correlation > 0.92 to prevent multicollinearity.
Reduced feature count from 32 → 23.
- Modeling & Evaluation
| Model | Train Accuracy | Test Accuracy | Precision | Recall | F1-score |
|---|---|---|---|---|---|
| Logistic Regression | 98.9% | 96.5% | 0.96–0.98 | 0.94–0.99 | 0.96 |
| KNN (default) | 96.7% | 95.6% | 0.94–0.98 | 0.91–0.99 | 0.95 |
| SVC (GridSearchCV) | ~100% | 97.4% | 0.96–1.00 | 0.96–0.99 | 0.97 |
.
🔧 Libraries Used python Copy Edit pandas, numpy, matplotlib, seaborn, sklearn, missingno 📈 Key Visualizations Class Distribution Plot
Feature Histograms
Heatmap for Feature Correlation
Missing Data Matrix using missingno
🧪 Model Evaluation Metrics Confusion Matrix
Accuracy Score
Classification Report
Precision
Recall
F1-score
✅ Conclusion All three models perform well, with SVC giving the highest accuracy and balance between precision/recall.
Feature selection based on correlation significantly improved generalization by reducing redundancy.
The project can be extended using:
Cross-validation
Ensemble models (e.g., Random Forest, XGBoost)
Feature engineering
SHAP/LIME for explainability