Final Exam = Data Mining = BA 360 ae Spring 2028 Lecturers L, BI Moubarki Hi, Bssghaier General Instructions: Read carefully each exercise, Circle the correct answer on the answer sheet (last page), Seratch paper is not allowed under any circumstances, Return only the answer sheet page # You have up to 120 minutes ‘© Questions can have multiple answers, «# Suspicion of cheating or using any unfair means of aid is enough to get your test withdrawn. |] «When you ate done, turn in the examination, Failure to do so will result in an automatic failing grade 1. Isit possible to design a logistic regression algorithm using a Neural Network Algorithm? A. TRUE B, FALSE Ist possible to apply a logistic regression algorithm on a 3-clas cassifiation problem? A TRUE B, FALSE ‘output to compare with target? A, AUG-ROC B, Accuracy ©. Specificityive Rate HVeRUDUDMOR True Posit OT 2 3 4 6 7 8 9 1 False Positive Rate Figute 1: ROC curves A. [00.1] B. [-00. +2] © 4) | D1} | 6. Suppose, You applied « Logistic Regression model on a given data and got a training accuracy «rl and testing accuracy az. Now, you want to add a few new observations to the same data, A, Training accuracy decreases B. Training accuracy increases or remains the same , Testing accuracy decreases D. Testing accuracy increases or remains the same Consider the ROC curves, illustrated in figure 1, of Three binary classifiers, 7. Which one of the three models (of the figure 1) is a random classifier A. Model 1 B. Model 2 C. Model 3 D, None of the above! Predicted Figure 2: Confusion Matrix: f- Ltrain, Xtest, y.train, ytest = train_test_split(X, y, test_size0.2, randon_state=42) g> model.fit(x train, y_train) ecole ecLopistichegressioNO i> print("Accuracy:", accuracy) > accuracy = accuracy score(y test, y_pred) 10. In Python programming language the argument "random.state® is used to: ‘A. makes a random experience repeatable B. makes the algorithm faster C. Fix the proportion of the RAM memory to be allocated D. Fix a maximum number of lines to display on the console 11. The Specificity index can be used to assess the ability of the model to detect true positive observations A. TRUE B. FALSE in the data. A. TRUE. B. FALSE Consider the confusion mati i15, Calculate the specificity of the class B (figure 2)? 16. Calculate the global accuracy (igure 2)? 17. The Out Of Bag Error generated by the random forest algorithm is: A. The percentage of observations inisclassified by the algorithm in the test set B. The percentage of observations misclassified by the algorithm in the training set. C. The average percentage of observations, which are misclassified, when the al- gorithm tests each single tree on the out of bag observations. 18. For Random Forest, the generated trees are independent of each other. A. TRUE B. FALSE 19. In case of unbalanced data, we can make a data augmentatiou, In such case. the test set should be generated before or after the data augmentation? A. Before B. After C. Does not matter 20. Bagging is the method for improving performance by aggregating the results of weak learners, A. TRUE B. FALSE 21. In the case of the Random Forest algorithm, you cau generate hundreds of trees (say 7. TaysTa) and then aggregate the results of these trees. Which of the following is true about individual(Tk) tree in Random Forest? A Individual tree is built on a subset of the features B. Individual tree is built on all the features C. Individual tree is built on a subset of observations: D. Individual tree is built on the full set of observations 22. How much is the entropy of the following set: Sa (=A t hits750 a 2 2 i Output layer Input k Hidden Layer ee fctvotion Functions with Softmax activation Relu function Figure 3: ANN Tt itemset TI {A,B CDE} T2| (AE) 13 | {E,D.C} Ta | {A.E,D.C} T5 | {A.B,C.E} Ts | {EAC} | {a,c} To) {GE} tio] {CEB} Table 1: Transaction Matrix 24. Consider the ANN architecture illustrated by the figure 3. The hidden layer contains two nodes with activation "Relu”. We use the "Softmax” activation function to predict the class label (A or B) of a given new observation. All needed weights and biases are given ‘on the graphic. Predict the class label of the observation (X1 = 1,X2=0.5,X3 = 038). A. Class A B. Class B C. Can not say 25. Consider the following transaction matrix, where A, B, C, D, E are 5 By using the table 1 calculate the confidence and the lift of t