Stampen

Question 1

K nearest neighbor requires three things

Answer 1

Set of stored records
Distance metric to compute distance
Value of k neighbors

Question 2

Advantages of KNN

Answer 2

Simple method
lack of parametric assumption
Good performance with large data sets

Question 3

Disadvantages of knn

Answer 3

Can be time consuming
Prohibits real time prediction of large dataset
Curse of dimensionality

Question 4

Advantages naive bayes

Answer 4

Computational efficient

Good classification prediction with many predictors

Question 5

Disadvantages of naive bayes

Answer 5

Requires large amount of records to obtain good results

Independence assumption may not hold for some variabele

Only categoriaal data

Question 6

Snowflake characteristics

Answer 6

Less restricted
Ability to store aggregations
Smaller data volumes

However,

Not easily understood by end users
High number of joins
Not predictable framework

Question 7

Dashboard conforms to three levels

Answer 7

Levels of perception
Comprehension of the current situation
Future status

Question 8

Pros decision

Answer 8

Easily understood

Easy to generate rules

Question 9

Cons decision trees

Answer 9

May suffer from overfitting
Does not handle correlated features well
can become quite large (needs pruning)
Does not handle streaming data easily

Question 10

When to stop expanding decision tree nodes

Answer 10

Stop expanding when all records have the same value
Stop expanding when all the records have similar attributes
stop if expansion does not improve inpurity measures
Watch out for overfitting, look for knee point

Question 11

4 ways to validate clusters

Answer 11

Cluster interpretability
Cluster stability (doesn't change when rows altered)
Cluster seperation (intra en inter good divided)
Number of clusters (useful number)

Question 12

Advantages of clustering

Answer 12

Does not require specifications of clusters
Purely data driven
Dendrograms easy to understand and interpret

Question 13

Limitations of clustering

Answer 13

Requires the computation and storage of nxn distance matrix
Makes only one pass through the data
Tends to have slow stability
Issue with respect to chosen distance metric
Hierarchical clustering sensitive to outliers

Question 14

Dimensions hierachies characteristics

Answer 14

Time related
unbalanced
Multiple branches of different types
Conforming dimensions