Transcript A Decision Tree Approach to Cube Construction

A Decision Tree Approach to
Cube Construction
Patrick Kelly
Data Cube Attributes
•
Designed for quick viewing and
decision-making
•
Built to optimize the users time
•
Cube construction is labor intensive
•
Users can only travel along pre-planned
path of data analysis
•
Dimensions are designed from preexisting warehouse tables
•
Reports are structured and ridges like
the construction of the cube
Decision Tree
Characteristics
• Decision Tree analysis is a
excellent analysis tool used in
data exploration
• Decision Trees are designed to
explicitly address the need to
identify the relationships
between the data’s variables.
• Decision Tree looks though more
relationships than the
multidimensional cube.
• Decision Tree verifies and
validates relationships as
statistically sound.
Comparison of Multidimensional Cubes
and Decision Trees
•
Multidimensional Cube
Decision Tree
Shows tabular views of data as tables with
relatively fixed dimensions; dimensions are
determined primarily on the basis of business
rules
Shows tabular views of data within relevant
dimensions as determined by computational
algorithms and business rules
Has database that is pre-built to support
anticipated queries
Has database that is pre-built to support
numerous unanticipated queries
Provides quick view retrieval
Has lengthy retrieval
Tends to limit number of cross-views or relevant
factors
Has few limitations on the relevant factors
Makes it difficult, almost impossible to identify
novel results
Emphasizes novel results and the identification
of important versus unimportant contributions
Pg 134 of Decision trees for Business Intelligence and Data Mining
Construction of Decision Trees Using Data
Cube Lixin Fu
• Decision Trees are not
suited for large
aggregated data sets
• Use a Sparse Statistics
Tree (SST) to design the
OLAP cube.
• Create Decision Tree
using the most current
data stored in the cube
Conclusion
• Decision Trees Decision
tree analysis is a good
analysis to be done in
data exploration
• Decision Tree analysis is
a possible validation
test for Cube