DTSA 5504 DATA MINING PIPELINE
COMPREHENSIVE EXAM 2026 ACTUAL
QUESTIONS AND CORRECT ANSWERS
GRADED A+
◉ Frequent Pattern Mining. Answer: closed pattern X = no super-pattern
Y given X with the same support; max-pattern X = no super-pattern Y
given X
◉ Apriori Algorithm. Answer: A fast method of finding frequent
itemsets, which also involves pruning non-frequent items and self-
joining of k-itemsets only if their first (k-1) items are the same.
◉ What challenges are there with apriori algorithm?. Answer: multiple
scans of whole dataset, huge number of candidates, support counting of
all candidates
◉ Improvements for apriori algorithm?. Answer: Partitioning, Sampling,
Transaction reduction
◉ Vertical data format. Answer: Mining frequent itemsets using the
________________________ is a method that transposes the rows of a
given data set into columns.
,◉ FP-Growth Algorithm. Answer: If 'd' is frequent in DB | abc, then
abcd is frequent (avoid candidate generation)
◉ Association Rules. Answer: Association rules specify a relation
between attributes that appears more frequently than expected if the
attributes were independent.
◉ Correlation rules. Answer: Measure of dependent/correlated events:
lift(A,B) = P(A U B) / P(A)P(B)
◉ Rules of lift (correlation). Answer: lift = 1 (independent), lift > 1
(positive), lift < 1 (negative)
◉ Metarule-Guided Mining. Answer: P1 and P2 and ..... and Px => Q1
and Q2 and .... and Qy
◉ Supervised Learning. Answer: Predefined classes, training data with
groundtruth label
◉ Unsupervised Learning. Answer: No predefined classes; aims to
identify potential clusters/patterns
◉ Classification. Answer: categorical class labels (e.g. fraud detection)
, ◉ Prediction. Answer: Continuous numerical values (e.g. stock prices)
◉ Steps of Classification. Answer: 1. Learning (Training Data, class
labels, model construction)
2. Classification (test data, model evaluation)
◉ What is considered in evaluation criteria?. Answer: accuracy, speed,
interpretability, robustness, scalability
◉ Decision Tree Induction. Answer: Basic algorithm: Attribute
selection, attribute split
Key properties: top-down, recursive (divide-and-conquer, greedy)
◉ Information Gain. Answer: Gain(A) = Info(D) - InfoA(D), where
InfoA(D) = sum(abs(Dj)/abs(D)) * Info(Dj) and Info(D) = -
sum(pi*log2(pi))
◉ Bayes' Theorem. Answer: The probability of an event occurring based
upon other event probabilities.
COMPREHENSIVE EXAM 2026 ACTUAL
QUESTIONS AND CORRECT ANSWERS
GRADED A+
◉ Frequent Pattern Mining. Answer: closed pattern X = no super-pattern
Y given X with the same support; max-pattern X = no super-pattern Y
given X
◉ Apriori Algorithm. Answer: A fast method of finding frequent
itemsets, which also involves pruning non-frequent items and self-
joining of k-itemsets only if their first (k-1) items are the same.
◉ What challenges are there with apriori algorithm?. Answer: multiple
scans of whole dataset, huge number of candidates, support counting of
all candidates
◉ Improvements for apriori algorithm?. Answer: Partitioning, Sampling,
Transaction reduction
◉ Vertical data format. Answer: Mining frequent itemsets using the
________________________ is a method that transposes the rows of a
given data set into columns.
,◉ FP-Growth Algorithm. Answer: If 'd' is frequent in DB | abc, then
abcd is frequent (avoid candidate generation)
◉ Association Rules. Answer: Association rules specify a relation
between attributes that appears more frequently than expected if the
attributes were independent.
◉ Correlation rules. Answer: Measure of dependent/correlated events:
lift(A,B) = P(A U B) / P(A)P(B)
◉ Rules of lift (correlation). Answer: lift = 1 (independent), lift > 1
(positive), lift < 1 (negative)
◉ Metarule-Guided Mining. Answer: P1 and P2 and ..... and Px => Q1
and Q2 and .... and Qy
◉ Supervised Learning. Answer: Predefined classes, training data with
groundtruth label
◉ Unsupervised Learning. Answer: No predefined classes; aims to
identify potential clusters/patterns
◉ Classification. Answer: categorical class labels (e.g. fraud detection)
, ◉ Prediction. Answer: Continuous numerical values (e.g. stock prices)
◉ Steps of Classification. Answer: 1. Learning (Training Data, class
labels, model construction)
2. Classification (test data, model evaluation)
◉ What is considered in evaluation criteria?. Answer: accuracy, speed,
interpretability, robustness, scalability
◉ Decision Tree Induction. Answer: Basic algorithm: Attribute
selection, attribute split
Key properties: top-down, recursive (divide-and-conquer, greedy)
◉ Information Gain. Answer: Gain(A) = Info(D) - InfoA(D), where
InfoA(D) = sum(abs(Dj)/abs(D)) * Info(Dj) and Info(D) = -
sum(pi*log2(pi))
◉ Bayes' Theorem. Answer: The probability of an event occurring based
upon other event probabilities.
