Please enable JavaScript.
Coggle requires JavaScript to display documents.
Chapter 3: Predictive Modeling (Modeling (Models (are created (for (a…
Chapter 3: Predictive Modeling
Modeling
Models
are created
for
a specific purpose
with
relevant information
to
uncover a target value
Related
using
feature vectors
fixed length
collection of
feature values
described by
set of attributes
tree structured
used to
create segments
addressing
multiple attributes
consists of
leaves
value for target variable
nodes
decision points
branches
connections between data
want
similarity within
difference between
Descriptive Modeling
uncovers
underlying phenomenon
Target
are
unknown interest
uncovered by
models
Induction
process of
creating model
from data
using
algorithms
from training data
Segmentation
Segmenting
process of
identifying
important attributes
based on
what says the most about target
supervised
considers
targets
how they will be effected
want
similarity
within segments
difference
between separate segments
Important Attributes
provide
most information about target
Information Gain
splitting criterion
on purity measure
entropy
#
Entropy
measures
disorder
in a data set
shows
variation in segment
based on
properties of interest
Visualizing
allows
introduction of
multiple dimensions
to the dataset
commonly
visualized with
scatterplots
Probability Estimation
used
because
provides
informative prediction
not
classification
Frequency Based
takes instances
within leaf nodes
and
applies to entire dataset
useful when
similarity between
different leaf nodes