CPS 521 - Introduction to Data Science final exam | Questions and Answers | Latest 2026 -

CPS 521 - Introduction to Data Science final exam | Questions and Answers | Latest 2026 - CPS521 Final Exam A Question 1: rA2 Question 2: elbow o? q Question 3: 12 3(' %5 Question 4: 4 2( Question 5 Random forest allows to generate hundreds of trees and then aggregate the results of these trees. Which of the following is true about an individual tree in random forest? () Itis built on a full set of features () Itis built on all available features (") None of the other options () Itis built on a full set of observations (@) It is built on a subset of features and observations Question 6 (1 point) Saved Which of the following is required by k-means clustering? () Number of clusters Initial guess of the cluster centroids ) Adefined distance metric (@) All of the above () None of the aboveQuestion 7 (1 point) Saved Which Python library and function is used to find the optimal value for a number of clusters in the K-Means algorithm? () Klbow from kneed () KneeLocator from sklearn (®) KneeLocator from kneed () None of other options Kneed from Kneelocator Question 8 (1 point) Saved A perceptron adds up all the weighted inputs it receives, and and sends it to the activation function that if sum exceeds a certain value, it outputs a 1, otherwise it just outputs a 0. What is the name of this activation function/feature in perceptron? (® Step function "VEU‘)Oan“* ’.L Hw QBCinpu’c‘LB f b] 0 () Perceptron function relorn 1L () None of other options e (¢ fetun @ () Logistic function (") Weighted functionGoogle What is back propagation? uestion 10 (1 point Saved What is back propagation Mcq? Q p ) What is back propagation? Explanation: Back propagation is the transmission of error back through i propag allow weights to be adjusted so that the network can learn. () None of the other options » ai-mcq-questions » neural-net... Neural Networks (Al) MCQ Questions & Answers - Letsfindcc () Another name for feed forwarding process Search for: What is back propagation Mcq? o . (_ ) Another name for the weighting process in perceptron What is back propagation how it works? () The transmission of errors through the network to allow for the Why is it called back propagation? - adjustment of inputs What is unsupervised learning method? = (®) The transmission of errors through the network to allow for the What are CNNs used for? adjustment of weights so that the network may learn MiiAacrtiam~ 1N 141 Question 9 (1 point) Saved Which of the the following is true about the perceptron training process? () Training consists of feeding it multiple training samples and calculating the — output for each of them / Weights are adjusted to minimize output error after each sample ~ ") Error rate of output is the difference between the desired outcome and the actual outcome - @ All of the above ) None of the above ©/ (W lnpukL) HWDinpuwt) *‘9] © relorn L e (¢ feluen @Not Tnput 1 only output wNov [ © Question . 11 (1 point) . Saved onlyPy Toput) L Which of the following perceptron desig gate? can be used to implement NOT logical Q A perceptron with two inputs with/weights wl= 1 and w2=1 respectively and bias parameter b= -0.5 O A perceptron with two inputs bias parameter b= 3 ith weights wl= -2 and w2=-2 respectively and O A perceptron with two inpyts with weights wl= 1 and w2=1 respectively and bias parameter b= -1.5 @ A perceptron with one input and weight w = -1 and bias parameter b = 0.5 o~ NS~ (") None of other options P .o a7 oa . FRY - Question 12 (1 point) Saved Which of the following perceptron designs can be used to implement AND logical gate? @ A perceptron with two inputs with weights wl= 1 and w2=1 respectively and bias parameter b=-1.5 :f:} A perceptron with two inputs with weights wl= -2 and w2=-2 respectively and bias parameter b= 3 () A perceptron with one input and weight w = -1 and bias parameter b = 0.5 () None of the other options Ifi:fi A perceptron with two inputs with weights wl= 1 and w2=1 respectively and bias parameter b= -0.5 (CEU.) inpq%f‘-) +@~’23('“P“*7'3 +b}o Tnputl | Trput 2 output AND O o ° ° refurn L ® ° | ° e ® , . . ferun @ ® ' [ 1 (LDGAPW L) 4 (W2 (nput2) +b = D+ MHM+ -1 S = 2-+S 0% HI S ReMNML * Reptot Fof PN @)Roubo A outpuk = cosrecA v/Toput! | Toput2 owput OR (DGAPWL) 4 (W2 (nput2) +b O+t 2 = (O + (DD - 05 @ ' ® i 2’0'5 @ ' v = -5 7)) Question 13 (1 point) Saving... & T~ Qervend / Which : of the following : perceptron designs . can be used to .implement OR logica X Repeat fos ald ol Rowx@ v owtput gate? ANS () None of the other options Co ) A perceptron with two inputs with weights wl= 1 and w2=1 respectively and bias parameter b= -0.5 (:) A perceptron with two inputs with weights wl= -2 and w2=-2 respectively and bias parameter b= 3 C) A perceptron with two inputs with weights wl= 1 and w2=1 respectively and bias parameter b= -1.5 (") A perceptron with one input and weight w = -1 and bias parameter b = 0.5 Question 14 (1 point) Saved Which of the following perceptron designs can be used to implement EXOR logical gate? Q A perceptron with one input and weight w = -1 and bias parameter b = 0.5 @} A perceptron with two inputs with weights wl= 1 and w2=1 respectively and bias parameter b= -1.5 Q A perceptron with two inputs with weights wl= 1 and w2=1 respectively and bias parameter b= -0.5 ’None of the other options N one Of th = AbOve O A perceptron with two inputs with weights wl= 1 and w2=1 respectively and bias parameter b= -1 4 NL Trput [ Trput2 CWPWr XOR (OGP L) 4 (WD Gnpwt2) +b (D) ) ° ° = MDY + -1 S - : ‘ - ® o . =15g OYIE ! o o Return A N7 doornd madch X . Repeat Sor oM @ Ny Wws norL of Houm m-l-oh . NV of H abuweQuestion 15 (1 point) Saved Which of the following statement IS TRUE about HO? The optimum hyperplane HO shown by the red line maximizes the margin The slop of the optimum hyperplane HO is equal to the slope of H+ ") The slop of the optimum hyperplane HO is equal to the slope of H- (® All of the above Queskon (6 Q16: What is the Slope of HO? 1 2 None of the above 0.71 1 ANSQuestion 17 (1 point) Saved One of the results of classifiers such as SVM for binary classification (N=2) is the confusion matrix as follows: TP - True Positive TN - True Negative FP - False Positive FN - False Negative TP fieo’%fun = |Predicted Values | Tp+6EF POSITIVE NEGATIVE F Rean[| = E— TP+ EN POSITIVE F -— I INEGATIVE Which of the following statements is correct calculation for accuracy? (") None of other options () (TP+TN)/(FP+FN) () TP/(TP+FN) (® (TP+TN)/Total () TP/ (TP+FP) ?reu fail R Qazaf/ Preeiston 4 &“‘7Question 18 (1 point) Saved Which of the following statement is true about using GD for this perceptron? MNe can not use GD for minimizing MSE as the cost function in this problem WW O We can use GD but the formula for adjusting w1 is wrong () We can use GD but the formula for adjusting w2 is wrong () We can use GD but using the 1 as the initial values of w1 and w2 is wrong (®) None of the other options Question 19 (1 point) Saved If we can use GD with the description above for this perceptron which of the followings are correct values in the next epoch? (@) wi= 0.56 and w2= 0.01 and y-cap= 42.1 and cost = 0.405 () wi=2.2 and w2= 3.7 and y-cap= 2.33 and cost = 0.054 Lpdets: L= (1~ W6 - bt ) (%" *2) (O wi1=2.2 and w2= 3.7 and y-cap= 42.1 and cost = 0.405 Weph . [~ é"”’"? X Grderts (}Pa(mf'é mfl () None of the other options () wi=0.56 and w2= 0.01 and y-cap= 233 and cost =0.054 ANSwer EQuestion 20 (1 point) Saved The minimum time complexity for training an SVM is O(n2). Based on that, what dataset sizes ARE NOT ideally suitable for SVM's? (") None of other options (@) Large datasets ~ (") Small datasets () Datasets of any size () Medium datasets Question 21 (1 point) Saved SVM's effectiveness is based on... () Choice of kernel ~ (_) 7N Presence of a soft margin. parameter '/ () Setting parameters for the kernel (®) Al of the above (") None of the aboveQuestion 22 (1 point) Saved Which of the following is the best description of a decision tree? (") None of other options O A graphical representation of a flow chart @ Flow-Chart & Structure in which internal node represents test on an attribute, each branch represents outcome of test and each leaf node represents class label () A flow-chart O A graphical representation of if rules Question 23 (1 point) Saved In building decision trees which of the following methods can be used to decide which attribute is better to split on as we descend the tree? (") Entropy v Q Information gain « () Gini g @ All of the above (") None of the above Q24:-maybe ANS= "anyone can see it’Questions 25 to 26 The next two question is related to the Jaccard similarity which is based on set theory and is used widely in text processing. It is defined as: SIM(A, B) = |A n B| / |A U B | which is a division of intersecting elements and a grand total of elements. Question 25 (1 point) Saved V2 324 5 6 7 ¥ 1 Considertwo setsA=%,2,4,5,8andB=1{1, 2,3, 5,6, 7, 9. . B e , ‘ What is the Jaccard similarity of Aand B~ 3ac_card S"Imar(fl = Totel Interseckion elements Q 0.55 Totald DIPE elements () 0.66 = 3/q =033 (") None of the other options (¢)0.33 ()0.45Jaccacd Similarity = Totel Interseckion elements Totald DIPE elements Question 26 (1 point) Saved Now consider the following two sentences: AI-|I'2J%1 : Hejs John :2_/ 0 + B: Heisa friend of Jonathan =0-2%5%F 4y 5 6 + What is the Jaccard similarityof AandB? (@) 0.2857 O None of the other options 004 ()0.7141 ()0.25 Question 27 (1 point) Saved Suppose a document D as a string of characters 1234124, and k = 2. Which one of the following is the 2-shingles set for the document D? Q None of the other options (® {12, 23,34,41, 12, 24} (){12,34,12,4"} Side Note (0 {1234, 4124} Lei’s say inskad we had k=3 e You Still mustoveriap eadn element (0)1{12, 23, 34, 41, 24} {223,234, 301,212 This Sizeshould be K=3Question 28 (1 point) Saved The following is the characteristic matrix for a collection of sets of words. Columns correspond to the sets; rows correspond to the components of universal set. Element S S S3 S4 m 0 1 0 1 0 1 0 | 0 p 0 1 0 0 q 1 0 1 1 n 1 0 1 1 When using jaccard similarity which of the following statements that compares the similarity score of the pair of documents is correct? () None of the other options SIM sy 6 L‘) - % @ SIM(S1, 54) = SIM(S3, S4) °1 (93/ 943 - ;; () SIM(S2, S4) = SIM(S3, 54) () SIM(S1, S2) = SIM(S1, S3) (®) SIM(S1, S3) = SIM(S1, S4) Question 29 (1 point) Saved Which of the following statements is correct? (") TF is the total number of times that a term occurs in a document / / |DF= 4 Q If a term appeared in all documents of a collection, its IDF is zero —2 MJ Z-? f"jz/ 2 Q If a term appeares in all documents of a collection its TF-IDF is zero 4y, Eoect @ All of the above (") None of the aboveQuestion 30 (1 point) Saved When finding communities, what is one important way that can be used to help narrow down the communities options by increasing the number of smaller communities? (::) Remove nodes with low node degrees @ Remove edges with the highest edge betweenness / () Remove edges with the lowest edge betweenness (") Remove nodes with high node degrees (") None of the other optionsm: © The first step of algorithm is to perform a breadth-first search (BFS) of the graph, starting at any of the nodes, making it a root. Here, we took G as the root node and found the following sub graph. Which of the following statement is correct about this graph shown below? Question 31 (1 point) ' Saved Dotled line means / D node is already visifed Fis difectly connected Yo D4E . Sinc& D ol ready -- - line O O OO O This graph is wrong because there is a direct connection between DF and AC edges O This graph is wrong because BFS is not used correctly (@) This graph is correct Q This graph will be correct by changing the DB edge solid line to a dashed line () None of the other optionsQuestion 32 (1 point) Saved The second step of the Girvan Newman algorithm is about the labeling of the nodes. We start by labeling the previous graph (by finding the correct form of that graph according to the first step) with G as the root by one, then we label each individual node by the sum of the labels of its parents. Which of the following statements is correct about the labelling results of the graph with G as the root when the second step is done correctly? () The label of G will be 2 after this step () The label of B will be 2 after this step P The label of E will be 2 after this step () All of the above (®) None of the aboveQuestion 33 (1 point) Saved Suppose after applying the third step and for the whole graph of the previous questions, we repeated the GN algorithm correctly when all nodes are considered the roots. Then we get the edge betweenness for the original graph with the following values that are shown below. Which of the following statements is correct when we want to find the communities in this graph? G o 'II:::I' If we choose the threshold of greater than 11 and remove edges with more betweenness than this threshold, then we get two communities of {A, B, C} and {D, E, F, G} .f::) If we choose the threshold of greater than 3 and remove edges with more betweenness than this threshold, then in the resulted communities, B is not connected to any node i::) If we choose the threshold of greater than 3 and remove edges with more betweenness than this threshold, then in the resulted communities, D is not connected to any node (@) All of the above () None of the above