[AI_107_A_2] 2. (3 pts each, 18 pts total)Consider a binary-class classification problem, where there are equal

numbers of positive and negative examples in the data set. Also suppose that the class label (i.e. +1)

is totally randomly assigned to these examples. Now for the following two classification models.

estimate their test accuracies based on different estimation methods.

(i) A perfect decision tree model trained from the training data without pruning.

1. Repeated random subsampling method, in which 2/3 of data for training, the rest for testing.

Test Accuracy = ?

2. Stratified 10-fold CV.

Test Accuracy = ?

3. LOOCV.

Test Accuracy = ?

(ii) A majority predictor model trained from the training data that always predicts the majority class in

training data.

1. Repeated random subsampling method, in which 2/3 of data for training, the rest for testing.

Test Accuracy = ?

2. Stratified 10-fold CV.

Test Accuracy = ?

3. LOOCV.

Test Accuracy = ?

2. (3 pts each, 18 pts total)Consider a binary-class classification problem, where there are equal numbers of positive and negative examples in the data set. Also suppose that the class label (i.e. + is totally randomly assigned to these examples. Now for the following two classification models, estimate their test accuracies based on different estimation methods. (i) A perfect decision tree model trained from the training data without pruning. (a) Repeated random subsampling method, in which 2/3 of data for training, the rest for testing. Test Accuracy = ? (b) Stratified 10-fold CV. Test Accuracy = ? (c) LOOCV. Test Accuracy = ? (ii) A majority predictor model trained from the training data that always predicts the majority class in training data. (a) Repeated random subsampling method, in which 2/3 of data for training, the rest for testing. Test Accuracy = ? (b) Stratified 10-fold CV. Test Accuracy = ? (c) LOOCV. Test Accuracy = ? 2. (3 pts each, 18 pts total)Consider a binary-class classification problem, where there are equal numbers of positive and negative examples in the data set. Also suppose that the class label (i.e. + is totally randomly assigned to these examples. Now for the following two classification models, estimate their test accuracies based on different estimation methods. (i) A perfect decision tree model trained from the training data without pruning. (a) Repeated random subsampling method, in which 2/3 of data for training, the rest for testing. Test Accuracy = ? (b) Stratified 10-fold CV. Test Accuracy = ? (c) LOOCV. Test Accuracy = ? (ii) A majority predictor model trained from the training data that always predicts the majority class in training data. (a) Repeated random subsampling method, in which 2/3 of data for training, the rest for testing. Test Accuracy = ? (b) Stratified 10-fold CV. Test Accuracy = ? (c) LOOCV. Test Accuracy =