Breast Cancer Classification using Support Vector Machines (SVM)

Submitted by: Tirtha Dutta
Model Type: Binary Classification | Supervised Learning

---

Objective

Use Support Vector Machines (SVM) to classify breast cancer tumors as Malignant (M) or Benign (B) using the Breast Cancer Wisconsin Dataset.

---

Dataset Info

• Source: Kaggle – Breast Cancer Dataset
• Total Records: 569 rows × 32 columns
• Target Variable: diagnosis (M = Malignant, B = Benign)

---

What You'll Learn

• Kernel-based classification (Linear & RBF SVM)
• Feature encoding and standardization
• Hyperparameter tuning using GridSearchCV
• Visualizing decision boundaries
• End-to-end ML workflow: from raw data to results

---

Workflow Summary

1. Data Cleaning & Preprocessing (Proof from Task 1)

• Dropped duplicate rows and the id column
• Verified missing values = 0
• Cleaned CSV saved as:
  data/breast_cancer_cleaned.csv

---

2. Exploratory Data Analysis (Proof from Task 2)

• Boxplot of features: images/eda_boxplot.png
• Heatmap of correlations: images/eda_heatmap.png

---

3. Feature Engineering (Proof from Tasks 1 & 6)

• Label encoded diagnosis:
  B → 0 (Benign) | M → 1 (Malignant)
• Scaled features using StandardScaler

---

4. Model Training & Hyperparameter Tuning

• Trained Linear SVM and RBF SVM
• Used GridSearchCV for tuning C and gamma
• Cross-validation (cv=5) accuracy: 97.89%
• Best parameters: C=10, gamma=0.01 (RBF Kernel)

---

5. Task 7 – Core Proofs

Proof Type	Output
Decision Boundary Plot	images/decision_boundary.png
Hyperparameter Tuning Heatmap	images/svm_rbf_tuning_heatmap.png
Classification Report	Printed in Notebook
Accuracy & CV Score	Printed via GridSearchCV

---

Final Model Results

Metric	Value
Accuracy	97.89%
Kernel	RBF
Best C	10
Best gamma	0.01

---

Tools & Libraries Used

• Python 3.10+
• Pandas, NumPy
• Matplotlib, Seaborn
• scikit-learn
• Jupyter Notebook

---

License

This project is licensed under the MIT License.