GoConqr hat ein neues Design! Sag uns, wie es dir gefällt.

Kostenlos registrieren

Karteikarten

BDMA

Leo Allert Deutsch 133 Karten

Karten in diesem Stapel (133)

  1. 11) Clustering and Text Mining are two of the data science methods discussed. To which dimensions of the CRISP-DM Cycle do these methods belong in particular? a) Business understanding and Data Understanding b) Business understanding, Data understanding, and Data Preparation c) Business understanding, Data Understanding, Data preparation, and modeling d) Business understanding, Data Understanding, Data preparation, modeling, and evaluationc) Business understanding, Data Understanding, Data preparation, and modeling
  2. 22) What is the Levenshtein metric between these two strings: Levenshtein metric Lavenstein metrik a) 3 b) 4 c) 5 d) 6a) 3
  3. 33) What is the sensitivity if the confusion matrix is TP=40 FP=20 FN=30 TN=10 a) 0,33 b) 0,25 c) 0,2 d) 0,5a) 0,33
  4. 45) Which of the following statements if FALSE about the ensemble techniques bagging and bootstrap aggregating (2 of them)? A: bagging draws without replacement n number of instances to create new subsets to train on the same model B: bagging reduces variance C: bootstrap aggregating is a variation of bagging that focusses on the wrong predictions and tries to improve wrong prediction by giving weight to data points D: bootstrap aggregating reduces varianceA: bagging draws without replacement n number of instances to create new subsets to train on the same model
    D: bootstrap aggregating reduces variance
  5. 56) Which statement is correct? I) Fitting a quadratic function with a linear model is an example of overfitting. II) Variance can be described as: a model that overfits is right on average, but is higly sensitive to specific training data. a) Only I is correct b) Only II is correct c) I & II are both correct d) I & II are both not correctb) Only II is correct
  6. 67) Which one of these ensemble methods applies classifiers sequentially, assigning higher weights to observations that have been mis-classified by the previous methods? a) Bagging b) Random Forest c) Boosting d) Bayesian Model Combinationc) Boosting
  7. 78) Which of the following is a desired property of a Big Data System? A) Minimal Maintenance B) High latency C) Flexibility D) AgilityA) Minimal Maintenance
  8. 89) What is not one of the 4 V's of big data? a) veracity b) variety c) visibility d) velocityc) visibility
  9. 910) Which statements are correct? (i) Entropy measures the general disorder of a dataset. (ii) Information gain measures the change in specificity due to any amount of new information being added. a) Both i and ii are correct. b) i is correct, ii is false. c) i is false, ii is correct. d) Both i and ii are false.b) i is correct, ii is false.
  10. 1011) Which statement about typical fitting graphs for tree induction is false? a) Generally, there will be more overfitting as one allows the model to be more complex. b) The 'sweet spot' indicates the point in the graph from where the holdout accuracy declines. c) The optimal tree size can be determined by looking at the highest training-set accuracy. d) Usually, the training-set accuracy is a little better than the holdout-set accuracy for each degree of complexity.c) The optimal tree size can be determined by looking at the highest training-set accuracy.
  11. 1112) Which statements are correct? (I) Regression Attempts to predict which of a (small) set of classes an individual belongs to. (II) Classification attempts to estimate or predict the numerical value of some variable for an individual a) I & II are true b) I & II are false c) I ist true; II is false d) I is false; II is trueb) I & II are false
  12. 1213) What does an entropy value of 0 mean? A. Minimum disorder B. No change in disorder C. Maximal disorder D. None of the above.A. Minimum disorder
  13. 1314) Which one of the following is not a layer in the Lambda Architecture. a. Batch Layer b. Access Layer c. Speed Layer d. Serving Layerb. Access Layer
  14. 1415) Which is the proper process for predictive modeling? a. Collect data > Define target > Build a model > Predict outcomes b. Collect data > Build a model > Define target > Predict outcomes c. Define target > Collect data > Build a model > Predict outcomes d. Build a model > Define target > Collect data > Predict outcomesc. Define target > Collect data > Build a model > Predict outcomes
  15. 1516) What is not a technique to avoid overfitting? a) Stop growing the tree before it gets too complex b) Grow the tree until it is too large, then "prune" it back c) Regularization d) Feature selectionc) Regularization
  16. 1617) Specialized tools used for web scrapping offer: a. More flexibility with less implementation efforts b. Less flexibility with less implementation efforts c. More flexibility with more implementation efforts d. Less flexibility with more implementation effortsb. Less flexibility with less implementation efforts
  17. 1718) Complexity of a tree induction model lies in a. Number of nodes b. Number of leafs c. Number of branches d. None of the abovea. Number of nodes
  18. 1819) Which properties desired in a Big Data System are not satisfied by the serving and batch layer? a) robustness b) scalability c) low latency d) debuggabilityc) low latency
  19. 1920) In which order occur the technologies on a typical Business Intelligence (BI) stack a) Data Sources  Data movement, streaming engines  Data warehouse servers  Mid-tier servers  Front-end applications b) Front-end applications  Data Sources  Data movement, streaming engines  Mid-tier servers - Data warehouse servers c) Front-end applications  Data movement, streaming engines  Data Sources  Mid-tier servers  Data warehouse servers d) Data Sources  Data warehouse servers  Data movement, streaming engines  Mid-tier servers  Front-end applicationsa) Data Sources  Data movement, streaming engines  Data warehouse servers  Mid-tier servers  Front-end applications
  20. 2021) Which method of data science tasks is an example of supervised method? A. Affinity grouping B. Similarity Matching C. Predictive Modeling D. ClusteringC. Predictive Modeling
  21. 2122) In the predictive modeling process, indicate the correct procedure A) Define target, collect data, build a model, predict outcomes B) Collect data, define target, build a model, predict outcomes C) Build a model, define target, collect data, predict outcomes D) Collect data, build a model, predict outcomes, evaluate outcomeA) Define target, collect data, build a model, predict outcomes
  22. 2223) Which is not one of the V’s of Big Data A) Volume B) Velocity C) Veracity D) ValidityD) Validity
  23. 2324) In Gartner Analytic Ascendancy Model, different types of analytics are explained based on the value and difficulty of each type. The correct sequence of information optimization is: A) Descriptive Analytics, Diagnostic Analytics, Predictive Analytics, and Prescriptive Analytics B) Diagnostic Analytics, Descriptive Analytics, Predictive Analytics, and Prescriptive Analytics C) Descriptive Analytics, Diagnostic Analytics, Prescriptive Analytics, and Predictive Analytics D) Diagnostic Analytics, Prescriptive Analytics, Descriptive Analytics, and Predictive AnalyticsA) Descriptive Analytics, Diagnostic Analytics, Predictive Analytics, and Prescriptive Analytics
  24. 2425) How can you describe privacy in a correct way? a. Ability to hide information about you, no one else needs to know. b. Ability to be the only one to use your identity c. Ability to have control over collection, use, disclosure and access of information about yourself d. Ability to decide on what happens to information about yourselfc. Ability to have control over collection, use, disclosure and access of information about yourself
  25. 2526) What use does the Jaccard distance have? a. To assess similarity between two data entries b. To assess dissimilarity between two data entries c. To asses similarity between multiple data entries d. To assess dissimilarity between multiple data entriesb. To assess dissimilarity between two data entries
  26. 2627) What does A ∪ B denote? a. Union: the region of A and B combined b. Union: the overlapping region of A and B c. Intersection: the region of A and B combined d. Intersection: the overlapping region of A and Ba. Union: the region of A and B combined
  27. 2728) What is not a desired property of a big data system? a. ad hoc queries b. low latency c. minimal maintenance d. distributedd. distributed
  28. 2829) What is NOT a desired property of a Big Data system? A. Robustness and fault tolerance B. High latency C. Minimal maintenance D. Ad hoc queriesB. High latency
  29. 2930) Which of the illustrated observations has on entropy value of 1? A. Yes, yes, yes, yes, yes, yes B. Yes, no, yes, no, yes yes C. Yes, yes, yes, no, no, no D. No, no, yes, no, no, noC. Yes, yes, yes, no, no, no
  30. 3031) The 4 V's consist of the following elements: a) Volume, Veracity, Velocity, Variety b) Veracity, Visibility, Velocity, Variety c) Variety, Veracity, Volume, Value d) Value, Visibility, Variability, Varietya) Volume, Veracity, Velocity, Variety
  31. 3132) What phase does NOT belong in the CRISP-DM model? a) Business Understanding b) Data Understanding c) Data Preparation d) Data Analysisd) Data Analysis
  32. 3233) What is Classification? A. It attempts to predict, for each individual in a population, which a (small) set of classes this individual belongs to. B. It attempts to estimate or predict, for each individual, the numerical value of some variable for that individual. C. It attempts to identify similar individuals based on data known about them. D. It attempts to find associations between entities based on transactions involving them.A. It attempts to predict, for each individual in a population, which a (small) set of classes this individual belongs to.
  33. 3334) Which statement(s) are correct? i). Regression attempts to estimate or predict the numerical value of some variable for an individual. ii). Classification attempts to predict which of a (small) set of classes an individual belongs to. A. (i) is true, (ii) is false B. (i) is false, (ii) is true C. (i) is true, (ii) is true D. (i) is false, (ii) is falseC. (i) is true, (ii) is true
  34. 3435) Which statement is false? (i) Entropy is zero at minimum disorder. (ii) Entropy is one at maximal disorder. (iii) Information gain measures how much an attribute increases entropy over the whole segmentation it creates. (iv) Information gain measures how much an attribute decreases entropy over the whole segmentation it creates. A. (i) B. (ii) C. (iii) D. (iv)C. (iii)
  35. 3536) According to the Gartner Analytic Ascendency Model, what is the correct order of analytics when analyzing a business problem? A: Descriptive Analytics > Diagnostic Analytics > Predictive Analytics > Prescriptive Analytics B: Diagnostic Analytics > Descriptive Analytics > Predictive Analytics > Prescriptive Analytics C: Prescriptive Analytics > Descriptive Analytics > Diagnostic Analytics > Predictive Analytics D: Prescriptive Analytics > Diagnostic Analytics > Descriptive Analytics > Predictive AnalyticsA: Descriptive Analytics > Diagnostic Analytics > Predictive Analytics > Prescriptive Analytics
  36. 3637) Which of the following statements about the Batch Layer (Lambda Architecture) is FALSE? A: Indexes the batch views so they can be queried with low latency. B: Runs in a loop and continuously recomputes the batch views from scratch. C: Scales by adding new machines. D: Manages the master data set.A: Indexes the batch views so they can be queried with low latency.
  37. 3738) Which of the following statements about Supervised Methods is TRUE? A: Is used for Causal (Explanatory) Modeling. B: Is used for Similarity Matching. C: Is used to cluster data points in (natural) groups. D: Is used for Profiling.A: Is used for Causal (Explanatory) Modeling.
  38. 3839) Which of the following is (are) the key benefits of MapReduce? A. Simplicity B. Scalability C. Speed D. All of the above.D. All of the above.
  39. 3940) Which of the following is (are) examples of ensemble method? A. Bagging (or bootstrap aggregation) B. Random forest C. Boosting D. All of the above.D. All of the above.
  40. 4041) Which statement is correct? I. Predictive modeling is the process of applying a statistical model or data mining algorithm to data to draw conclusions about past observations. II. Explanatory modeling is the use of statistical models for explaining how the world works (by testing causal explanations) A. Statement I is true B. Statement II is true C. Both statements are true D. Both statements are wrongB. Statement II is true
  41. 4142) Which distance measure is most appropriate when comparing objects/persons described by (many) binary variables? A. Euclidian distance B. Manhattan distance C. Jaccard distance D. Canberra distanceB. Manhattan distance
  42. 4243) In the Lambda Architecture, which problem of the Batch Layer, does the Speed Layer overcome? a. The Speed Layer has more up-to-date information as it complements information in real-time. b. The Speed Layer uses a more advanced technology which is quicker than the technology used in the Batch Layer. c. The Speed Layer recomputes all data from scratch, to make sure it is more complete. d. None of the above.a. The Speed Layer has more up-to-date information as it complements information in real-time.
  43. 4344) Which of the following is NOT a desired property of Big Data systems as listed by Marz (2015)? a) Scalability b) Extractability c) Generalisation D) Extensibilityb) Extractability
  44. 4445) Which statement is correct: i) Term counts within the document form the Term Frequency values for each term. ii) The document counts across the corpus form the Inverse Document Frequency value. a) i) is correct, ii) is incorrect. b) i) is correct, ii) is correct. c) i) is incorrect, ii) is correct. d) i) is incorrect, ii) is incorrect.b) i) is correct, ii) is correct.
  45. 4546) Which characteristic is not a difficulty in analyzing text? A. Too much text B. Messy C. Stopwords D. StemmingA. Too much text
  46. 4647) What is the MapReduce tool useful for? a. To reduce the number of leaves in a decision tree. b. To process data in the batch layer of the Lambda Architecture. c. To reduce the number of maps required to find a location. d. To merge the batch layer and the speed layer to a serving layer.b. To process data in the batch layer of the Lambda Architecture.
  47. 4748) Which of the following statements is correct? 1. High entropy means there is a high disorder between properties, which is good for our data science problem. 2. High information gain is good for our data science problem. A) Both statements are correct B) 1 is correct, 2 is incorrect C) 1 is incorrect, 2 is correct D) Both statements are incorrectB) 1 is correct, 2 is incorrect
  48. 4849) In the Nearest Neighbors for prediction (kNN), what will happen if one has k=1? A) Overfitting B) Underfitting C) Nothing D) Depends on the specific caseA) Overfitting
  49. 4950) From which of the following systems can we collect data for a Data Mining project? A. Data Movement, Streaming Engines B. Data Warehouse Servers C. Mid-Tier Servers D. All of the aboveD. All of the above
  50. 5051) What is the most valued skill for a data scientist? A. evaluation B. data understanding C. feature engineering D. modellingC. feature engineering
  51. 5152) What does entropy measure? A. change in dataset disorder B. average of variable C. information gain dataset disorder D. dataset disorderD. dataset disorder
  52. 5253) What does the expected value framework do? A. convert a ROC curve into a profit curve B. convert a profit curve into a ROC curve C. convert a AUC curve into a profit curve D. convert a profit curve into a ROC curveA. convert a ROC curve into a profit curve
  53. 5354) Which of the following models could be used for both supervised and unsupervised data-mining? [A] Clustering [B] Class probability estimation [C] Similarity matching [D] Regression[C] Similarity matching
  54. 5455) Which of the following is not a desired property of a Big Data system? [A] Ad hoc queries [B] Low latency [C] Robustness [D] Maximal maintenance[D] Maximal maintenance
  55. 5556) What is the simplest method to limit three size? [A] Define the least amount of information gain that must be present at each split [B] Define the least number of instances that must be present in a node [C] Cut off branches and replace them with leaf nodes [D] Prune a tree if its result does not reduce accuracy on the test set[B] Define the least number of instances that must be present in a node
  56. 5657) Which of the following statements is true about the speed layer and the batch layer in the Lambda architecture? A. Queries from users can be run directly on the batch layer but not on the speed layer. B. The speed layer presents data through the serving layer, which the batch layer does not do. C. The batch layer looks at all the data at once whereas the speed layer only looks at recent data. D. Output from the speed layer comes per table, output from the batch layer comes per batch.C. The batch layer looks at all the data at once whereas the speed layer only looks at recent data.
  57. 5758) Which of the following is NOT an example of ensembling? A. Bagging B. Cross-Validation C. Boosting D. BlendingB. Cross-Validation
  58. 5859) Which of the following is NOT a source fir the differences between predictive and explanatory modeling? A. Correlation-association B. Theory-data C. Retrospective-prospective D. Bias-varianceA. Correlation-association
  59. 5960) What is NOT a characteristic of the batch layer in the Lambda Architecture Framework A. The master data is immutable B. It runs in a loop and continously recomputates C. It only looks at recent data D. It pre-computes query functions, called batch viewsC. It only looks at recent data
  60. 6061) Which statement is correct? 1. Web scraping through a modern programming language offers a lack of flexibility but requires less effort. 2. Web scraping through specialized tools allow faster implementation and provide more flexibility What is true about these two statements? A. Statement 1 is true, statement 2 is false B. Statement 1 is false, statement 2 is true C. Both statements are true D. Both statements are falseD. Both statements are false
  61. 6162) In which stages of the CRISP-DM cycle is Clustering useful? Select the answer that best answers this question. A. Clustering is useful in the Data Understanding and Data Preparation stage. B. Clustering is useful in the Business Understanding and Data Understanding stage. C. Clustering is useful all throughout the CRISP-DM cycle. D. Clustering is useful all throughout the CRISP-DM cycle, but especially in the Business Understanding and Data Understanding stage.D. Clustering is useful all throughout the CRISP-DM cycle, but especially in the Business Understanding and Data Understanding stage.
  62. 6263) Which statement is correct? 1- A model that overfits is right on average, but is highly sensitive to specific training data. 2- The regression model can be used for both supervised and unsupervised learning. A. Statement 1 is true, statement 2 is false B. Statement 1 is false, statement 2 is true C. Both statements are true D. Both statements are falseA. Statement 1 is true, statement 2 is false
  63. 6364) Which of the following statements about train and test sets is correct? 1. Having smaller test sets leads to higher variances across folds 2. Having a larger trainset leads to better performance in each model" A. Statement 1 is true, statement 2 is false B. Statement 1 is false, statement 2 is true C. Statement 1&2 are true D. Statement 1&2 are falseC. Statement 1&2 are true
  64. 6465) In chapter 1 by Marz & Warren (2015), they discuss a number of techniques when attempting to scale up with a traditional database. What technique was NOT mentioned? A. Queueing: Batch requests by adding in a queue between the web server and the database B. Sharding: use multiple databases and spread the table across all servers C. Merging: combining multiple data points into one row in the table which is then uploaded D. None of the aboveC. Merging: combining multiple data points into one row in the table which is then uploaded
  65. 6566) According to BDMA Session 1, the 4 V’s of Big Data are: 1) Variety 2) Velocity 3) Volume and 4) Veracity. Which of the following statements is correct? A. The Variety of big data has always been used as a source for innovation B. The input data Velocity is not well understood C. Volume of big data is the best understood D. The main attention of researchers is on VeracityC. Volume of big data is the best understood
  66. 6667) Which of the following descriptions for methods of gathering data relates to web scrapping? A. Manually browsing a site and copy-pasting from it B. Using protocols to interact with specific web sites that can be used by any registered user C. Using a programming language or specialized tool to browse a website and collect data from it D. Paying to a firm to collect web data for youC. Using a programming language or specialized tool to browse a website and collect data from it
  67. 6768) According to Provost and Fawcett, which statement about information gain is incorrect? A. Information gain is a measure of how informative an attribute is with respect to the target B. Information gain is based on a purity measure called entropy C. Information gain measures how much an attribute improves entropy over the whole segmentation it creates D. Information gain measures the change in entropy due to any amount of incorrect information being removedD. Information gain measures the change in entropy due to any amount of incorrect information being removed
  68. 6869) What is the correct order of the data mining process (without taking into account iteration)? A. Business understanding, Data understanding, Data preparation, Modeling, Evaluation, Deployment B. Data understanding, Business understanding, Data preparation, Modeling, Evaluation, Deployment C. Business understanding, Data understanding, Data preparation, Evaluation, Modeling, Deployment D. Business understanding, Data preparation, Data understanding, Modeling, Evaluation, DeploymentA. Business understanding, Data understanding, Data preparation, Modeling, Evaluation, Deployment
  69. 6970) Which of the following corresponds with a higher number of folds? a) Having to run more models, having smaller train sets and and smaller test sets b) Having to run more models, having larger train sets and smaller test sets c) Having to run more models, having larger train sets and larger test sets d) Having to run less models, having larger train sets and smaller test setsb) Having to run more models, having larger train sets and smaller test sets
  70. 7071) Which of the following abilities is not an ability that a good data science manager should possess according to Provost & Fawcett (2013). A. Be able to anticipate outcomes of data science projects B. Be able to produce proof-of-concept studies as these studies are highly predictive for the outcome of a project C. Be able to communicate well with and be respected by both “techies” and “suits” D. Be able to understand and appreciate the needs of the business in order to interact with their counterparts in other functional areas of the company.B. Be able to produce proof-of-concept studies as these studies are highly predictive for the outcome of a project
  71. 7172) Suppose that you want to predict what news stories will cause a significant change in stock price. You would like predictive text mining to recommend news stories that you should pay attention to in order to make good investments. Which is not one of the problems that you will come across? A. It is difficult to predict exactly what the stock price will be B. It is difficult to associate a specific piece of news with a price change C. It is difficult to predict large changes in stock price, so instead we predict relatively small changes D. It is difficult to predict the effect of the news story far in advanceC. It is difficult to predict large changes in stock price, so instead we predict relatively small changes
  72. 7273) Which of the following represents the 4V's of Big Data? A. Volume, Velocity, Veracity, Variety B. Volume, Velocity, Veracity, Vulnerability C. Volume, Velocity, Vulnerability, Variety D. Volume, Vulnerability, Veracity, VarietyA. Volume, Velocity, Veracity, Variety
  73. 7374) Suppose you are building a linear model that needs to model complex regularities with large training sets. What model best fits these requirements? A. Support Vector Machine B. Tree Induction C. Neural Networks D. Linear RegressionB. Tree Induction
  74. 7475) Which statement about entropy is true? A. Entropy is 0 at minimum disorder. B. Entropy is 1 at minimum disorder. C. Entropy is 0 at maximum disorder. D. Entropy is 0.5 at maximum disorder.A. Entropy is 0 at minimum disorder.
  75. 7576) In a fitting graph, what is the reason for the curve of holdout accuracy? A. The curve keeps improving when a model becomes more complex, because it is better able to place all datapoints. B. The curve is concave, because a simple model underfits and a complex model overfits C. The curve is convex, because simple models overfit and complex models underfit D. The curve keeps improving because a model gets better with more available data, like depicted in the learning curveB. The curve is concave, because a simple model underfits and a complex model overfits
  76. 7677) What term involves principles, processes, and techniques for understanding phenomena via the (automated) analysis of data. A. Data science B. Data strategy C. Data engineering D. Data miningA. Data science
  77. 7778) Which layer of the lambda architecture stores the master copy of the dataset and precomputes views on that master data set? A. Serving Layer B. Speed Layer C. Storing Layer D. Batch LayerD. Batch Layer
  78. 7879) What is on the x-axis and y-axis of a ROC curve? A. x-axis is true positive rate and y-axis is false positive rate B. x-axis is fraction of test instances and y-axis is accuracy C. x-axis is false positive rate and y-axis is true positive rate D. x-axis is accuracy and y-axis is percentage of test instancesC. x-axis is false positive rate and y-axis is true positive rate
  79. 7980) Which of the following concepts is NOT linked to Deep Learning? A) Artificial Neural Networks B) Supervised Pattern Recognition C) Backpropagation Algorithm D) Machine LearningB) Supervised Pattern Recognition
  80. 8081) According to the lecture by Rodrigo Belo and Otto Koppius, which of the statements below is true about crossvalidation? A) Cross-validation is a more sophisticated training and testing procedure B) Cross-validation computes estimates over all the data by performing multiple splits and systematically swapping out samples for testing C) Cross-validation estimates generalization performance D) All of the aboveD) All of the above
  81. 8182) How is the complexity of tree model measured? a. by number of parameters b.by number of classes c. by number of nodes d. number of treesc. by number of nodes
  82. 8283) There is no trade off between: a. error rate and accuracy b. true positive and true negative c. overfitting and generalisationb. true positive and true negative
  83. 8384) In which case have you overfit a tree induction? A: When each leaf node has one single instance B: When accuracy outperforms AUC C: When accuracy on your test set is 100% D: When AUC outperforms accuracyA: When each leaf node has one single instance
  84. 8485) Which of the following options below is not considered a key property of data? A) Perpetuity B) Rawness C) Transparency D) ImmutabilityC) Transparency
  85. 8586) Which of the following statements is true about explanatory modeling? 1. Explanatory models rely on associations between measurable variables 2. Explanatory modeling seeks to minimize the combination of model bias and sampling variance A) Statement 1 & 2 are both true B) Only statement 1 is true C) Only statement 2 is true D) Statement 1 & 2 are both not trueD) Statement 1 & 2 are both not true
  86. 8687) Which of the following attempts to find associations between entities based on transactions involving them? A. classification B. co-occurrence grouping C. a & b D. none of the aboveB. co-occurrence grouping
  87. 8788) Which of the following is NOT true for both when comparing CRISP-DM and a software development project? a) The project might change scale and scope as elements become more clear. b) The involvement of key stakeholders in the problem definition phases is critical. c) Even iterations that do not produce any meaningful solutions can still be valuable. d) None of the aboved) None of the above
  88. 8889) The errors that a model makes can be characterized by which of the below factors? A - Inherent Randomness; B - Bias; C - Variance A. A & B B. B & C C. A & B & C D. C onlyC. A & B & C
  89. 8990) In the Lambda architecture, which layer will have the most interaction with the master dataset? A) Speed layer B) Query layer C) Serving layer D) Batch layerD) Batch layer
  90. 9091) Which of the following statements is not true regarding overfitting? A. Increasing the amount of minimum tree size lowers the chance of overfitting B. There is no general analytic way to avoid overfitting C. One can create a validation set (nested hold-out set) out of the training set in order to avoid overfitting on the test set D. All of the above statements are correctA. Increasing the amount of minimum tree size lowers the chance of overfitting
  91. 9192) What statement regarding Lambda architecture is NOT correct? A. Lambda Architecture is relatively resilient to human errors B. Lambda Architecture splits two layers into a final, serving layer C. Lambda Architecture provides low latency D. Lambda Architecture consists of a batch, speed and serving layer.B. Lambda Architecture splits two layers into a final, serving layer
  92. 9293) Which of these statements about cross-validation is correct? I. A higher number of folds will lead to a lower number of leaves in a decision tree. II. 5 folds will usually have a lower variance across models than 10 folds on the same training set. A) I is correct, II is incorrect. B) I is incorrect, II is correct. C) Both I and II are correct. D) Both I and II are incorrect.B) I is incorrect, II is correct.
  93. 9394) Which is NOT a desired property of a big data system? A) Robustness and fault tolerance B) High latency C) Minimal Maintenance D) Ad hoc queriesB) High latency
  94. 9495) Which of the following options contains four properties of a Big Data system? A. Ad hoc queries, Scalability, Privacy, Robustness B. Scalability, Robustness, Manageability, Low latency reads, and updates C. Ad hoc queries, Robustness, Extensibility, Debuggability D. Scalability, Extensibility, Robustness, ComplexityC. Ad hoc queries, Robustness, Extensibility, Debuggability
  95. 9596) Little complexity of a decision tree model can be achieved by forcing the model to have a) a small minimal leaf size b) a small number of nodes c) a small number of parameters d) all of the aboved) all of the above
  96. 9697) When applying the k-nearest neighbor's algorithm, using k=1 will result in: A) A very complex model, but not very useful B) A very complex model, but likely to overfit C) A very simple model, but not very usefully D) A very simple model, but likely to overfitB) A very complex model, but likely to overfit
  97. 9798) Select the correct answer I: Unsupervised data mining methods are best applicable when there is a specific target variable II: Clustering is an example of an unsupervised data mining method A) I is correct, II is incorrect. B) I is incorrect, II is correct. C) Both I and II are correct. D) Both I and II are incorrect.B) I is incorrect, II is correct.
  98. 9899) Which of the following options represents the 6 steps of the CRISP-DM cycle? a. Business understanding, Data understanding, Data analysis, Modelling, Evaluation and Deployment b. Business understanding, Data understanding, Data preparation, Modelling, Evaluation and Deployment c. Business understanding, Data understanding, Data preparation, Modelling, Monitoring and Deployment d. Business understanding, Data assessment, Data preparation, Modelling, Evaluation and Deploymentb. Business understanding, Data understanding, Data preparation, Modelling, Evaluation and Deployment
  99. 99100) Which type of Gartner's Analytic Ascendancy Model requires the least human input? a. Predictive analysis b. Descriptive analysis c. Diagnostic analysis d. Prescriptive analysisa. Predictive analysis
  100. 100101) Modeling: I Clustering is an unsupervised method. II Profiling is a supervised method. III Predictive modeling is a supervised method. a) I is correct. b) I and II are correct. c) I and III are correct. d) All of the answers are correct.c) I and III are correct.
  101. 101102) Which of the following statements regarding performance is FALSE? a) With smaller data, tree induction tends to overfit more. b) Tree induction is more flexible for larger training sets. c) For smaller training set sizes, logistic regression yields better generalization accuracy. d) Generalization performance improves as more test data are available.d) Generalization performance improves as more test data are available.
  102. 102103) Which of the following options describes the four aspects of the broad definition of privacy? a. 1. Invasions into protected spaces, relationships or decisions 2. Collection of information 3. Use of information 4. Disclosure of information b. 1. Invasions into protected spaces, relationships or decisions 2. Collection of information 3. Analyzing information 4. Disclosure of information c. 1. Invasions into protected spaces, relationships or decisions 2. Collection of information 3. Use of information 4. Making commercial use of information d. 1. Invasions into personal information 2. Collection of information 3. Use of information 4. Disclosure of informationa. 1. Invasions into protected spaces, relationships or decisions 2. Collection of information 3. Use of information 4. Disclosure of information
  103. 103104) Among the desired properties of a big data system, one of them relates to the fact that systems need to behave correctly despite challenges such as machines going down, duplicated data, human errors etc. Which one is it? A) Robustness and fault tolerance B) High latency C) Minimal maintenance D) Ad hoc queriesA) Robustness and fault tolerance
  104. 104105) What of the following does predictive modeling seek to minimize? 1. model bias 2. Standard deviation A. only 1 B. only 2 C. Both 1&2 D. Neither of themA. only 1
  105. 105106) Which of the following statement is true? 1. Overfitting is the tendency to tailor models to the training data 2. The more complex a model is, the less likely it will overfit" A. only 1 B. only 2 C. Both 1&2 D. Neither of themA. only 1
  106. 106107) Assuming a constant computing speed, Every recalculation in the batch layer A. Takes as long as the previous calculation B. Takes longer than the previous one, if new data has been collected in the meanwhile C. Takes as long as the previous calculation, if there is a speed layer that takes care of the new data D. b and c are correctB. Takes longer than the previous one, if new data has been collected in the meanwhile
  107. 107108) " y = ß0 + ß1x1 + ß2x2 + u" What is the meaning of "u" in this formula? a. parameter to estimate b. explanatory variable c. dependent variable d. error termd. error term
  108. 108109) What is not an element of the CRISP-DM framework? A. Data understanding B. Data preparation C. Modelling D. InterpretationD. Interpretation
  109. 109110) Which of the following is not an issue regarding NN methods? (A) Intelligibility (B) Dimensionality (C) Complexity (D) Computational efficiency(C) Complexity
  110. 110111) Which of the following is not a way to avoid overfitting on a Tree Induction? (A) Reduce the number of variables (B) Specify a minimum number of instances in a leaf (C) Specify a minimum amount for the information gain at each split (D) Prune the tree(A) Reduce the number of variables
  111. 111112) What is overfitting? A. When the models are tailor made optimized to fit the training sets B. When the data are optimized to provide accuracy C. When the models have similar performances on the training set and the test set D. When the learning curve of the model is linear with additional dataA. When the models are tailor made optimized to fit the training sets
  112. 112114) How can an ROC curve be converted into a profit curve? a. By maximising the profit for the optimal scenario. b. By plugging in information about the benefits and costs of each scenario. c. By establishing a base scenario to assume class priors. d. It cannot be converted into a profit curve.b. By plugging in information about the benefits and costs of each scenario.
  113. 113115) What are the relationships between entities called in a fact based model? a) Nodes b) Properties c) Centroids d) Edgesd) Edges
  114. 114116) What are/is the advantage(s) of a fact based model to the relational database? a) Data is raw b) The data is eternally true c) Has normalized and denormalized data d) All of the abovec) Has normalized and denormalized data
  115. 115117) Which model tries to fit the fattest bar when discriminating for classes? a) Linear regression b) Logistic regression c) Neural network d) support vector machined) support vector machine
  116. 116119) Which of the following is NOT true? A. Entropy is -1 at minimum disorder B. Error rate is in inverse proportion to accuracy. C. The process of recursively segmenting the population stops when the maximum allowed tree depth has been reached D. Logistic regression can also be used for classificationA. Entropy is -1 at minimum disorder
  117. 117120) What is a definition of 'data mining'? A. the extraction of knowledge from data, via technologies that incorporate the principles from data science B. the technologies used to store and process Big Data C. the practice of basing decisions on data instead of pure intuition D. the strategy for creating and capturing value from dataA. the extraction of knowledge from data, via technologies that incorporate the principles from data science
  118. 118121) Which of the following statement about Inverse Document Frequency (IDF) is correct? A. a normal word will always have high IDF B. a unique word will always have high IDF C. IDF is a property of a term in the document D. none of the aboveB. a unique word will always have high IDF
  119. 119122) Which is correct? Statement 1: Explanatory models are based on underlying causal relationships between constructs while predictive models rely on associations between variables. Statement 2: Explanatory modelling seeks to minimize the combination of model bias and sampling variance, predictive modelling seeks to minimize model bias to obtain the most accurate representation of the underlying model." A) Both statements are true B) Statement 1 is true and statement 2 is false C) Statement 2 is true and statement 1 is false D) Both statements are falseC) Statement 2 is true and statement 1 is false
  120. 120123) Which is correct? Statement 1: K-means clustering is a supervised learning method. Statement 2: In K-means clustering, the actual solution depends on the starting centroid choice. A) Both statements are true B) Statement 1 is true and statement 2 is false C) Statement 2 is true and statement 1 is false D) Both statements are falseC) Statement 2 is true and statement 1 is false
  121. 121124) Which statement is false? A) Each node in the tree contains a test of an attribute B) Complexity is a measure of the flexibility of a model C) Each leaf in a decision tree contains a value for the target variable D) Direction and the size of document are both important in cosine similarityD) Direction and the size of document are both important in cosine similarity
  122. 122125) Which is not one of the difficulties when analysing text? A) Frequency B) Messy C) Synonyms D) StemmingA) Frequency
  123. 123126) Which of the following statements is true about data mining projects? I. Data mining projects are to be managed like software engineering projects with focus on performance optimization. II. In data mining projects, iterations are about approaches and strategy. A) Only I is correct B) Only II is correct C) Both I and II are correct D) None of above is correctB) Only II is correct
  124. 124127) Which step/process is practical implementation of data science? a. Data mining b. Data engineering c. Data-driven decision making d. Data strategya. Data mining
  125. 125128) Which parts of the CRIS-DM cycle do you find ethical dilemmas about data collection? a. Business understanding, Data and Data preparation. b. Data understanding, Data and Data preparation. c. Business understanding, Data and Data understanding. d. Data preparation, Data and Deployment.c. Business understanding, Data and Data understanding.
  126. 126129) Which of the following options is not an aspect of privacy? a. Search of information b. Collection of information c. Hiding of information d. None of the abovec. Hiding of information
  127. 127130) Which of the following methods has a specific target variable? a. Similarity matching b. Clustering c. Affinity grouping d. Causal modelingd. Causal modeling
  128. 128131) A bagging method causes a ... A) Lower variance B) Lower bias C) Lower variance and lower bias D) None of the aboveA) Lower variance
  129. 129132) A stacking ensemble method is to ... A) Select the best performing model B) Boost lower performing weights C) Combine different models to one model D) Combine multiple datasetsD) Combine multiple datasets
  130. 130133) A low K value in nearest neighbour A Predict the value closer to the average of the dataset B Predicts an instance by a larger amount of close instances C Normalizes the variable D Predicts an instance by a smaller amount of close instancesD Predicts an instance by a smaller amount of close instances
  131. 131134) What kind of problem is not typically solved with a supervised method: A. classification B. Regression C. Causal modelling D. similarity matchingD. similarity matching
  132. 132135) Which of the following is NO measure of similarity between two entities? A. Edit Distance B. Manhattan Distance C. Chicago Distance D. Jaccard DistanceC. Chicago Distance
  133. 133136) Which is not one of the V’s of Big Data A) Volume B) Velocity C) Veracity D) ValidityD) Validity

Informationen

Beschreibung
Karteikarten am BDMA, erstellt von Leo Allert am 05/12/2022.
Erstellt von
Leo Allert
Aufrufe
2
Bewertung

Feedback

Worum geht es?

Wie gefällt dir das neue Design?

Dein Eindruck