Suppose you want to learn a decision tree from the simple dataset listed in the table...

     

Transcribed Image Text:

Suppose you want to learn a decision tree from the simple dataset listed in the table below. There are 8 samples, each with 3 binary attributes (X1, X2, X3) and a binary class label (Y). Use the data to answer the following questions. Instance 1 X, X Y 0 0 0 0 2 0 0 1 0 3 0 1 I 6 7 8 (a) Compute the entropy of the class label. (b) Calculate the information gain for each of the three attributes. (c) Which attribute should be selected for the root of the decision tree? Why? (d) After the root node is selected, the entire tree can be learned by recursively splitting the data into two subgroups, finding the next best attribute to split on, dividing the subgroup into smaller groups, and so forth. How do you know when to stop growing the tree? In other words, what are the stopping criteria? (e) By manually running the algorithm described above, draw the resulting tree. (f) Compute the training error. (g) Now suppose you are presented new instances for which the class Y is unknown. Use your decision tree to predict the label of each instance listed below. Instance X X, X Y 9 1 1 1 ? 10 1 0 0 ? 11 0 1 1 ? (h) Do you have any basis on which to evaluate if the tree is overfitting? Why or why not? How might you combat overfitting in a decision tree? Suppose you want to learn a decision tree from the simple dataset listed in the table below. There are 8 samples, each with 3 binary attributes (X1, X2, X3) and a binary class label (Y). Use the data to answer the following questions. Instance 1 X, X Y 0 0 0 0 2 0 0 1 0 3 0 1 I 6 7 8 (a) Compute the entropy of the class label. (b) Calculate the information gain for each of the three attributes. (c) Which attribute should be selected for the root of the decision tree? Why? (d) After the root node is selected, the entire tree can be learned by recursively splitting the data into two subgroups, finding the next best attribute to split on, dividing the subgroup into smaller groups, and so forth. How do you know when to stop growing the tree? In other words, what are the stopping criteria? (e) By manually running the algorithm described above, draw the resulting tree. (f) Compute the training error. (g) Now suppose you are presented new instances for which the class Y is unknown. Use your decision tree to predict the label of each instance listed below. Instance X X, X Y 9 1 1 1 ? 10 1 0 0 ? 11 0 1 1 ? (h) Do you have any basis on which to evaluate if the tree is overfitting? Why or why not? How might you combat overfitting in a decision tree?

Answer rating: 100% (QA)

To answer the questions regarding the decision tree based on the provided dataset lets go through each question step by step a To compute the entropy of the class label we need to calculate the entrop... View the full answer