Best Model: The SVC with RBF kernel achieved the highest accuracy (98.25%) and F1-score (0.98). The RBF kernel effectively captures non-linear decision boundaries in the breast cancer dataset, leading to superior performance.
Training Time: Linear models (LinearSVC, SVC_linear) trained faster due to simpler optimization, while the polynomial kernel in LinearSVC_poly increased computational complexity. The RBF kernel, though slightly slower than linear models, balanced speed and performance.
Regression Comparison
Table 2: Regression Models Performance
Model
Training Time (s)
MSE
RMSE
R² Score
SVR_rbf
1.1617
0.3743
0.6118
0.7130
SVR_linear
0.2364
0.4640
0.6812
0.6442
SVR_poly
3.9859
1.0626
1.0308
0.1852
Justification:
Best Model: The SVR with RBF kernel outperformed others with the lowest MSE (0.3743) and highest R² score (0.7130), indicating strong predictive capability for housing prices.
Training Time: The linear kernel was fastest but less accurate. The polynomial kernel performed poorly, likely due to overfitting or suboptimal hyperparameters (e.g., degree=2). The RBF kernel, while slower than linear, provided the best trade-off.
Regression Model Visualization
Figure 1: 3D Surface Plot of SVR_rbf Predictions
Visualization of the RBF-based SVR model predicting housing prices using two key features (e.g., MedInc and HouseAge). The red surface represents model predictions, while blue points are training data.
Figure 2: Contour Plot of SVR_rbf Predictions
Contour lines show predicted housing price regions. Darker colors indicate higher prices, aligning with feature relationships.
Conclusion
Classification: Use SVC_rbf for high-accuracy tumor diagnosis.
Regression: SVR_rbf is ideal for housing price prediction due to its balance of speed and accuracy.
Plots demonstrate the RBF model’s ability to capture complex patterns in regression tasks.
Loading Document
