AN 300 EXAM 2 UPDATED ACTUAL QUESTIONS AND
CORRECT ANSWERS
Question:
Descriptive Analysis
Answer:
Question: *"What has happened" * Techniques: *Data Mining* Tools: *JMP
PRO* Application: -Summarize financial performances -Summarize
purchasing behavior
Question:
Cluster Analysis is..
Answer:
Used to discover *natural grouping* of objects -Objects within a group is
similar -Objects across groups is dissimilar
Question:
Cluster Analysis problem definition:
Answer:
Giving data on objects of interest ->Find the # of groups and group
memberships Organize objects into groups -Maximize similarities of objects
within a group -Minimize similarities of objects between groups
Question:
Cluster Analysis similarity measures...
Answer:
Euclidean distance for numerical data -Straight line -Standardization
Matching coefficient for categorical data "Number of columns within matching
values/ total number of columns"
Question:
Cluster Analysis Business Applications:
Answer:
1. Market segmentation- *target* marketing
2. Personalization- *financial* advising
3. Quality control- *outlier detection*
Question:
Cluster Analysis methods evaluation:
Answer:
Cubic Clustering Criterion (CCC) -CCC>2: *good fit*
-0<=CCC <= 2: *possible fit* -CCC~-30: *presence of outliers*
Question:
Association Rule is:
Answer:
used to discover *nontrivial "what goes with what" connections* among
groups of items in distinct events or transactions.
, Question:
Association Rule problem definition
Answer:
Given data on the co-occurrences of items of interest -> Find the likelihood of
the co-occurrences of these items Express as IF-THEN statements: -IF item-set:
*condition* -THEN item-set: *Consequent*
Question:
Association Rile evaluation
Answer:
-Support: Proportion of occurrence -A high value -> high frequency of
occurrence -Support (X->Y)= (# of events including both X and Y)/(Total # of
events)
Question:
Cont'd
Answer:
Confidence- accuracy -A high value-> THEN item-set occurs frequently with
the IF item-set -Confidence (X->Y)= (#of events including X and Y) / (# of events
including only X)
Question:
Cont'd
Answer:
Lift-strength over random guesses -*Greater than 1 -> better than guessing*
-Lift (X -> Y)=
(# of events including X and Y *times* Total # of events) / (# of events including
only X *times* Number of events including only Y)
Question:
Association Rule business applications:
Answer:
1. Products placement
2. Medical diagnosis
3. Website navigation optimization
4. Quality control
Question:
Text Mining is...
Answer:
used to *discover trends and patterns* from text data in a collection of
documents
Question:
Text Mining problem definition
Answer:
Given data on objects of interest in word, phrases, sentences, and paragraphs
-> Find patterns and/or trends in textual data -Visualize results as a word cloud
-Analyze results: ~Cluster analysis (latent class analysis) ~Association rule
analysis
CORRECT ANSWERS
Question:
Descriptive Analysis
Answer:
Question: *"What has happened" * Techniques: *Data Mining* Tools: *JMP
PRO* Application: -Summarize financial performances -Summarize
purchasing behavior
Question:
Cluster Analysis is..
Answer:
Used to discover *natural grouping* of objects -Objects within a group is
similar -Objects across groups is dissimilar
Question:
Cluster Analysis problem definition:
Answer:
Giving data on objects of interest ->Find the # of groups and group
memberships Organize objects into groups -Maximize similarities of objects
within a group -Minimize similarities of objects between groups
Question:
Cluster Analysis similarity measures...
Answer:
Euclidean distance for numerical data -Straight line -Standardization
Matching coefficient for categorical data "Number of columns within matching
values/ total number of columns"
Question:
Cluster Analysis Business Applications:
Answer:
1. Market segmentation- *target* marketing
2. Personalization- *financial* advising
3. Quality control- *outlier detection*
Question:
Cluster Analysis methods evaluation:
Answer:
Cubic Clustering Criterion (CCC) -CCC>2: *good fit*
-0<=CCC <= 2: *possible fit* -CCC~-30: *presence of outliers*
Question:
Association Rule is:
Answer:
used to discover *nontrivial "what goes with what" connections* among
groups of items in distinct events or transactions.
, Question:
Association Rule problem definition
Answer:
Given data on the co-occurrences of items of interest -> Find the likelihood of
the co-occurrences of these items Express as IF-THEN statements: -IF item-set:
*condition* -THEN item-set: *Consequent*
Question:
Association Rile evaluation
Answer:
-Support: Proportion of occurrence -A high value -> high frequency of
occurrence -Support (X->Y)= (# of events including both X and Y)/(Total # of
events)
Question:
Cont'd
Answer:
Confidence- accuracy -A high value-> THEN item-set occurs frequently with
the IF item-set -Confidence (X->Y)= (#of events including X and Y) / (# of events
including only X)
Question:
Cont'd
Answer:
Lift-strength over random guesses -*Greater than 1 -> better than guessing*
-Lift (X -> Y)=
(# of events including X and Y *times* Total # of events) / (# of events including
only X *times* Number of events including only Y)
Question:
Association Rule business applications:
Answer:
1. Products placement
2. Medical diagnosis
3. Website navigation optimization
4. Quality control
Question:
Text Mining is...
Answer:
used to *discover trends and patterns* from text data in a collection of
documents
Question:
Text Mining problem definition
Answer:
Given data on objects of interest in word, phrases, sentences, and paragraphs
-> Find patterns and/or trends in textual data -Visualize results as a word cloud
-Analyze results: ~Cluster analysis (latent class analysis) ~Association rule
analysis
