chapter 4

Question 1

What are the three different types of data mining tasks that can be used depending on the business need?

Answer 1

Prediction, clustering, or association.

Question 2

What is classification in data mining prediction?

Answer 2

what is being Predicted is a class label (e.g., weather: sunny, cloudy, rainy).

Question 3

What is regression in data mining prediction?

Answer 3

Predicting a numeric value (e.g., temperature: 31).

Question 4

What kind of learning does classification employ?

Answer 4

Supervised learning.

Question 5

What is the most frequently used data mining method?

Answer 5

Classification.

Question 6

What is nominal data?

Answer 6

Data labelled/classified into mutually exclusive categories within a variable.

Question 7

What is ordinal data?

Answer 7

Statistical data where variables have natural, ordered categories.

Question 8

What are the two steps in classification-type prediction?

Answer 8

Model development/training and model testing/deployment.

Question 9

What is the simple split method in classification?

Answer 9

Splitting data into 2 mutaally exclusive sets
(eg: training (~70%) and testing (30%) sets.)

Question 10

What is k-fold cross-validation?

Answer 10

Data split into k subsets; k training/testing experiments are done.

Question 11

What is a confusion matrix used for?

Answer 11

Accuracy estimation in classification problems.

Question 12

What is a True Positive (TP)?

Answer 12

A correctly predicted positive case.

Question 13

What is a False Positive (FP)?

Answer 13

An incorrectly predicted positive case.

Question 14

What is a False Negative (FN)?

Answer 14

An incorrectly predicted negative case.

Question 15

What is a True Negative (TN)?

Answer 15

A correctly predicted negative case.

Question 16

What are classification techniques in data mining?

Answer 16

Classification techniques are methods used to assign data items to predefined classes based on their attributes.

Question 17

List the classification techniques mentioned in the slides.

Answer 17

Decision tree analysis, statistical analysis, neural networks, support vector machines, case-based reasoning, Bayesian classifiers, genetic algorithms, and rough sets.

Question 18

True or False: Decision tree analysis is a type of classification technique.

Answer 18

True.

Question 19

True or False: Neural networks are included under classification techniques.

Answer 19

True.

Question 20

Which classification technique is based on probabilistic reasoning?

Answer 20

Bayesian classifiers.

Question 21

Which classification technique mimics biological evolution concepts?

Answer 21

Genetic algorithms.

Question 22

Which classification technique focuses on similarity to past cases?

Answer 22

Case-based reasoning.

Question 23

Which classification technique separates classes using optimal boundaries?

Answer 23

Support vector machines.

Question 24

What is a decision tree?

Answer 24

A decision tree is a classification method that employs a divide-and-conquer approach to classify data.

Question 25

What method do decision trees employ?

Answer 25

Decision trees employ a divide-and-conquer method.

Question 26

How does a decision tree divide a training set?

Answer 26

It recursively divides a training set until each division consists of examples from one class.

Question 27

What happens when decision tree division is complete?

Answer 27

Each division contains examples from only one class.

Question 28

What is the first step in building a decision tree?

Answer 28

Create a root node and assign all of the training data to it.

Question 29

What is the second step in building a decision tree?

Answer 29

Select the best splitting attribute.

Question 30

What does “best splitting attribute” mean in decision trees?

Answer 30

It refers to the attribute that best separates the data into distinct classes based on a chosen splitting criterion.

Question 31

What is the third step in building a decision tree?

Answer 31

Add a branch to the root node for each value of the split.

Question 32

What does splitting the data into mutually exclusive subsets mean?

Answer 32

Each data instance belongs to only one subset created by the split.

Question 33

What is done after creating branches in a decision tree?

Answer 33

The data is split into mutually exclusive subsets along the specific split.

Question 34

What is the fourth step in building a decision tree?

Answer 34

Repeat steps 2 and 3 for each and every leaf node until the stopping criteria is reached.

Question 35

When does the decision tree building process stop?

Answer 35

When the stopping criteria is reached.

Question 36

What are the main aspects in which decision tree algorithms differ?

Answer 36

Splitting criteria, stopping criteria, and pruning.

Question 37

What does splitting criteria refer to in decision tree algorithms?

Answer 37

It refers to which variable is selected and what value is used to split the data.

Question 38

What does stopping criteria determine in decision trees?

Answer 38

It determines when to stop building the tree.

Question 39

What is pruning in decision trees?

Answer 39

A generalization method that determines which parts of the tree to remove.

Question 40

Why is pruning used in decision trees?

Answer 40

To improve generalization and reduce overfitting.

Question 41

List the most popular decision tree algorithms.

Answer 41

ID3, C4.5, C5, CART, CHAID, and M5.

Question 42

True or False: CART is a decision tree algorithm.

Answer 42

True.

Question 43

True or False: CHAID is used in decision tree construction.

Answer 43

True.

Question 44

What is cluster analysis used for in data mining?

Answer 44

Automatic identification of natural groupings of things.

Question 45

Which learning family does cluster analysis belong to?

Answer 45

The machine-learning family.

Question 46

What type of learning does cluster analysis employ?

Answer 46

Unsupervised learning.

Question 47

What does unsupervised learning mean in clustering?

Answer 47

Learning patterns from data without a predefined output or target variable.

Question 48

True or False: Cluster analysis has an output variable.

Answer 48

False.

Question 49

How does cluster analysis handle new instances?

Answer 49

It learns clusters from past data and then assigns new instances to those clusters.

Question 50

What is cluster analysis called in marketing?

Answer 50

Segmentation.

Question 51

What can clustering results be used to identify?

Answer 51

Natural groupings of customers.

Question 52

How can clustering help in targeting or diagnostic purposes?

Answer 52

By identifying rules for assigning new cases to classes.

Question 53

What role does clustering play in population analysis?

Answer 53

It provides characterization, definition, and labeling of populations.

Question 54

True or false: Clusting decreases the size and complexity of problems for other data mining methods.

Answer 54

True

Question 55

How does clustering help with rare-event detection?

Answer 55

By identifying outliers in a specific domain.

Question 56

What statistical methods are used in cluster analysis?

Answer 56

k-means, k-modes, and similar statistical methods.

Question 57

Which neural network methods are used in cluster analysis?

Answer 57

Adaptive Resonance Theory (ART) and Self-Organizing Maps (SOM).

Question 58

What role does fuzzy logic play in clustering?

Answer 58

It enables fuzzy clustering such as the fuzzy c-means algorithm.

Question 59

Which evolutionary technique is used in clustering?

Answer 59

Genetic algorithms.

Question 60

What is an important question to answer before clustering?

Answer 60

How many clusters?

Question 61

What is k-means clustering?

Answer 61

A clustering algorithm that partitions data into k pre-determined clusters.

Question 62

What does “k” represent in k-means clustering?

Answer 62

The pre-determined number of clusters.

Question 63

What is Step 0 in the k-means clustering algorithm?

Answer 63

Determine the value of k.

Question 64

What is Step 1 in the k-means algorithm?

Answer 64

Randomly generate k random points as initial cluster centers.

Question 65

What is Step 2 in the k-means algorithm?

Answer 65

Assign each point to the nearest cluster center.

Question 66

What is Step 3 in the k-means algorithm?

Answer 66

Re-compute the new cluster centers.

Question 67

What is the repetition step in k-means clustering?

Answer 67

Repeat steps 2 and 3 until a convergence criterion is met.

Question 68

What does convergence usually mean in k-means clustering?

Answer 68

The assignment of points to clusters becomes stable.

Question 69

What is association rule mining?

Answer 69

A data mining method that finds interesting relationships (affinities) between variables.

Question 70

True or False: Association rule mining is popular in business.

Answer 70

True.

Question 71

What types of relationships does association rule mining discover?

Answer 71

Relationships between items or events.

Question 72

Which learning family does association rule mining belong to?

Answer 72

The machine-learning family.

Question 73

What type of learning does association rule mining use?

Answer 73

Unsupervised learning.

Question 74

Does association rule mining have an output variable?

Answer 74

No, there is no output variable.

Question 75

What is another name for association rule mining?

Answer 75

Market basket analysis or affinity analysis.

Question 76

List business applications of association rule mining.

Answer 76

Cross-marketing, cross-selling, store design, catalog design, e-commerce site design, optimization of online advertising, product pricing, and sales/promotion configuration.

Question 77

How is association rule mining used in medicine?

Answer 77

To identify relationships between symptoms and illnesses, diagnoses and patient characteristics, treatments in medical DSS, and genes and their functions.

Question 78

What is the generic form of an association rule?

Answer 78

X ⇒ Y [S%, C%].

Question 79

What do X and Y represent in an association rule?

Answer 79

Products and/or services.

Question 80

What is X called in an association rule?

Answer 80

The left-hand-side (LHS) or antecedent.

Question 81

What is Y called in an association rule?

Answer 81

The right-hand-side (RHS) or consequent.

Question 82

What does support (S) measure in association rules?

Answer 82

How often X and Y occur together.

Question 83

Write the formula for support of an association rule.

Answer 83

Supp(X→Y) = (Number of baskets that contain both X and Y) / Total number of baskets.

Question 84

What does confidence (C) measure in association rules?

Answer 84

How often Y occurs given X.

Question 85

Write the formula for confidence of an association rule.

Answer 85

Confidence(X→Y) = Supp(X→Y) / Supp(X).

Question 86

True or False: All association rules are interesting and useful.

Answer 86

False.

Question 87

Why are association rule algorithms needed?

Answer 87

To discover and identify association rules from data.

Question 88

List algorithms used for discovering association rules.

Answer 88

Apriori, Eclat, FP-Growth, and their derivatives and hybrids.

Question 89

What do association rule algorithms first identify?

Answer 89

Frequent item sets.

Question 90

What happens to frequent item sets after identification?

Answer 90

They are converted into association rules.

Question 91

What is the Apriori algorithm?

Answer 91

An algorithm that finds subsets common to at least a minimum number of itemsets based on support.

Question 92

What criterion does Apriori use to find frequent subsets?

Answer 92

Minimum support value.

Question 93

What approach does the Apriori algorithm use?

Answer 93

A bottom-up approach.

Question 94

How does Apriori expand frequent subsets?

Answer 94

One item at a time, increasing subset size incrementally.

Question 95

What is the progression of subset sizes in Apriori?

Answer 95

One-item subsets, then two-item subsets, then three-item subsets, and so on.

Question 96

How are candidate groups evaluated in Apriori?

Answer 96

They are tested against the data for minimum support value.