DSCI 4520 - DATA MINING EXAM I REVIEW
QUESTIONS WITH VERIFIED ANSWERS
Dimension - Answers - ________ refers to the # of input variables (degrees of freedom).
Over-fitting - Answers - ________ is caused by an overly complex model that is too
flexible; can also mean a situation in which there are too many variables in your model.
"Which split is best?" - Answers - ___________ refers to the Splitting Criteria method.
Splitting Criteria methods - Answers - The ___________ that are most widely used are
based on Pearson Chi-Square test, Gini index, and entropy; all three measure
difference in class distributions across the chid nodes and usually end with similar
results.
Pure child node - Answers - If a split ends in a __________, the split is undisputedly
best.
Parent node - Answers - If the expected target is same in child nodes as in the
____________ then the split is WORTHLESS.
Pruning - Answers - _______ creates a sequence of trees of increasing complexibility.
Assessment criterion - Answers - _______ is needed for deciding the best (sub) tree.
Benefits of trees - Answers - _______ include interpretability, mixed measurement
scales, and missing values.
Interpretability - Answers - _______ is a tree-structured presentation.
Binary outcome - Answers - We CANNOT predict a __________.
Binary - Answers - When a response variable is ______, the data set (from past
information) are zeroes.
Simple Linear Regression - Answers - A model with one predictor variable is called
__________.
Residual - Answers - The _______ is the unexplained part of Y in the Simple Linear
Regression formula.
Dummy - Answers - Indicator variables are also known as _________ variables.
QUESTIONS WITH VERIFIED ANSWERS
Dimension - Answers - ________ refers to the # of input variables (degrees of freedom).
Over-fitting - Answers - ________ is caused by an overly complex model that is too
flexible; can also mean a situation in which there are too many variables in your model.
"Which split is best?" - Answers - ___________ refers to the Splitting Criteria method.
Splitting Criteria methods - Answers - The ___________ that are most widely used are
based on Pearson Chi-Square test, Gini index, and entropy; all three measure
difference in class distributions across the chid nodes and usually end with similar
results.
Pure child node - Answers - If a split ends in a __________, the split is undisputedly
best.
Parent node - Answers - If the expected target is same in child nodes as in the
____________ then the split is WORTHLESS.
Pruning - Answers - _______ creates a sequence of trees of increasing complexibility.
Assessment criterion - Answers - _______ is needed for deciding the best (sub) tree.
Benefits of trees - Answers - _______ include interpretability, mixed measurement
scales, and missing values.
Interpretability - Answers - _______ is a tree-structured presentation.
Binary outcome - Answers - We CANNOT predict a __________.
Binary - Answers - When a response variable is ______, the data set (from past
information) are zeroes.
Simple Linear Regression - Answers - A model with one predictor variable is called
__________.
Residual - Answers - The _______ is the unexplained part of Y in the Simple Linear
Regression formula.
Dummy - Answers - Indicator variables are also known as _________ variables.
