Exam AWS Certified Machine Learning - Specialty topic 1 question 93 discussion

Due to straight angles, I would choose Decision tree. See https://scikit-learn.org/stable/auto_examples/classification/plot_classifier_comparison.html#sphx-glr-auto-examples-classification-plot-classifier-comparison-py

B - Decision tree - is not the best answer. If you use decision tree to do clustering, every time you need to partition the space into 2 parts. Hence you will split the space into 3*3. The red points in the center box and the black points will fall into the 8 boxes around it. The black points will be identified as 8 different classes. C is the correct answer. SVM with non-linear kernel is appropriate for non-linear clustering. Even if the shape is close to rectangular. SVM with non-linear kernel will be ale to approximate the rectangular boundary shape.

Tip: When details are missing, assume ideal conditions, so assume no overfitting issues. Therefore is B is better than C. If there are overfitting issues or a posibility of overfitting then C is the right answer.

Decision tree makes more sense, this decision boundary isn't complex at all and there is no risk of overfitting, all the points are inside the square

This is because the RBF kernel can handle non-linear relationships between features, which is often necessary for complex classification tasks.

Decision Tree can treat the training data well but will have a risk of overfitting. the SVM with RBF kernel will be more robust.

As the positive cases can be interpreted and separated from non positive ones by decision tree easily. SVM would have made sense if the two classes were inseparable or had complex relationship in data.

It is C SVM with RBF Kernel can classify this image. For decision tree, it will be more difficult

From the visual information provided, an SVM with an RBF kernel (Option C) would likely be the best choice because it can handle the circular class distribution. The RBF kernel is especially good at dealing with such scenarios where the boundary between classes is not linear.

Answer C B. Decision Tree: Decision trees can capture non-linear patterns and are capable of splitting the feature space in complex ways. They can be very effective if the decision boundary is not linear, but they might also overfit if the decision boundary is too complex. C. SVM with RBF Kernel: An SVM with a radial basis function (RBF) kernel is designed to handle non-linear boundaries by mapping input features into higher-dimensional spaces where the classes are more likely to be separated by a hyperplane. Given the clustered nature of the classes in the image, an SVM with an RBF kernel would likely be able to separate the classes with a higher degree of accuracy.

SVM-RBF is the correct solution

Support vector machine (SVM) with a radial basis function kernel would likely have the highest accuracy for this task because it can handle the non-linear separation required by the data.

I will lean with C

Answer is B as Decision tree can attain 100% accuracy in this case.

SVM with RBF and proper C and Gamma value can accomodate this square shape (https://vitalflux.com/svm-rbf-kernel-parameters-code-sample/)

confusing between SVN or devision tree. learning towards C

Answer C In general, SVMs are a good choice for tasks where accuracy is critical, such as fraud detection and medical diagnosis. Decision trees are a good choice for tasks where interpretability is important, such as customer segmentation and product recommendation.