Data mining Final
Which of the following weights are not typically initialized to random values?
-0.25
Which of the following are general guidelines for choosing a network architecture?
A number of hidden layers, size of hidden layers, number of output nodes
Which of the following algorithms can be used for basket analysis?
A priori, Frequent pattern growth
With sufficient iterations, neural net can easily overfit the data. To avoid overfitting all of the following should be done EXCEPT:
Add nodes to capture complexity
Which of the following are characteristics of Average Distance?
Also called average linkage, Distance between two clusters is the average of all possible pair-wise distances, Most popular with Centroid distance measurement
All of the following are examples of data mining paradigms EXCEPT:
Analysis
(blank) is the study of what goes with what.
Association Rules
Benefits of Market Basket Analysis are
Better location of items in store to promaote impulse buys, Selection of items for joint promotions and marketing programs, Enable categorization of shopper purpose and motivations
Which of the following are benefits of Market Basket Analysis?
Better location of items, Joint promotions and market programs, Enables categorization
Which of the following options are part of the network structure of neural nets?
Bias values, Weights, nodes, multiple layers.
Big errors lead to ___ changes in weights
Big
The center of a class is known as the:
Centroid
Which of the following are data mining paradigms?
Classification, prediction, association
Discriminant analysis was used for which of the following before data mining?
Classifying organisms into species, skulls, and fingerprint analysis all used discriminant analysis for classification long before data mining.
which of the following is not a classification technique for discriminant analysis before data mining was invented?
Cluster analysis
What is the percentage of antecedent transactions that also have the consequent item set?
Confidence
_________ shows the rate at which consequents will be found
Confidence
. A _________ is a visual representation of the cluster hierarchy.
Dendrogram
A net with a single output node and several hidden layers, where g is the identity function, take the same form as a linear regression model
False
Bias values are subject to iterative adjustment
False
Classification Rules are also called market basket analysis.
False
Confidence is the number or percentage of times this rule occured over the total transactions.
False
Confidence is the number or percentage of times this rule occurred over the total transactions.
False
Discriminant Analysis is best suited for large data sets.
False
During the initial pass through a network, the error is propagated back and distributed to the first hidden node and used to update its weight
False
For batch updating all records in the training set are fed to the network after updating takes place
False
For the terms, the "IF" part is known as the consequent and the "THEN" part is know as the antecedent.
False
In a neural net structure, model "coefficients" are tweaked only a few times.
False
In case updating, completion of all records through the network is one source.
False
It is best to use classes with unequal frequencies in your data set
False
Market business analysis uses the frequency of single objects to suggest business rules.
False
Support is the support for the rule over the support for the antecedent.
False
The 3 layers of the network structure are the input, middle, and output layer.
False
The confidence of a rule can be calculated without first knowing the support.
False
The first step to algorithm for discrimination analysis is to create classification score that reflects the distance from each class?
False
The goal of clustering is to form groups of different records
False
The goal of the Initial Pass through the Network is to find weights that yield the best errors.
False
The lift ration shows the rate at which consequents will be found.
False
True or false: A neural network contains multiple layers; these layers include the input layer, the output layer, and the weight layer.
False
he most successful applications in data mining of neural networks have been multilayer feed-backward networks.
False
In order to measure lift, you must subtract the probability of the outcome from the confidence.
False.
True or False.Some disadvantages of neural networks include that is has good predictive ability, it can capture complex relationships, and there is no need to specify a model.
False.
Which is NOT part of the Multiple Layers
Horizontal Layer
All of the following are layers in the network structure besides:
Known Layers
To prevent overfitting the data which of the following should be done?
Limit the number of training epochs, do not overtrain data, examine the performance on the validation set.
Cluster analysis can be used for which of the following?
Market Segmentation, Industry Analysis, Market Structure Analysis
In maximum distance clustering or complete linkage, the distance between two clusters is the __________ distance between the two pair of records.
Maximum
Which of the following are steps to algorithm for discriminant analysis?
Measuring distance, Classification functions, converting to probabilities.
Which of the following is an example of cluster analysis?
Periodic table of elements, classification of species, grouping securities in portfolios
What is the goal of the initial pass through network?
To find weights that yield best predictions
A lift value greater than 1 indicates that a rule is useful in finding consequent item sets.
True
A neural network is a model for classification and prediction.
True
Association rules, or Affinity analysis, constitute a study of "what goes with what."
True
Dependent Variable uses categorical variables
True
Discriminant Analysis is a classical statistical technique that was used for classification long before data mining.
True
In a neural network, the error associated with the weights begins to decrease due to thousands of updates being performed
True
In case updating weights are updated after each record id ran through the network
True
Inter-cluster distance are maximized between clusters
True
Lift is another metric used to determine the most interesting rules
True
Neural Networks are known to be "black boxes."
True
Small errors leave weights relatively unchanged
True
Step 2 in discriminant analysis is classification functions.
True
The goal of cluster analysis is to form groups of of similar records
True
The idea behind neural networks is to combine the input information in a very flexible way that captures complicated relationships among these variables and between them and the response variable.
True
The most successful applications in data mining of neural networks have multi-layer feed-forward networks.
True
The parable of "beer and diapers" has proven there is a trend of when customers buy diapers they also buy beer.
True
Training the model means estimating the weights that lead to the best possible predictive results.
True
When using logistic regression you can use either categorical variables and interval variables.
True
clustering is used for segmenting markets into groups of similar customers
True
The following are all types of layers in a network structure except:
Weight layer
A net with a single output node and no hidden layers, where g is the identity function, takes the same form as
a linear regression model
Confidence is the percent of ________ transactions that also have the consequent item set
antecedent
The center of a class is called a(n):
centeroid
Discriminant analysis tends to be considered more of a ____________ method than a data mining method.
statistical classification
Which of the following are Association Rules?
study of "what goes into what", transaction based or event based
Using the Euclidean distance has ____ drawbacks?
three
Which criteria should be used to stop updating the neural network?
when the weight change is negligible, the miss-classifications rate reaches a required threshold, when a limit on runs in reached
