Database Reference
In-Depth Information
TreeGrowing (
S
,
A
,
y
,
SplitCriterion
,
StoppingCriterion
)
Where:
S
- Training Set
A
- Input Feature Set
y
- Target Feature
SplitCriterion
--- the method for evaluating a certain split
StoppingCriterion
--- the criteria to stop the growing process
Create a new tree
T
with a single root node.
IF
StoppingCriterion
(
S
)
THEN
Mark
T
as a leaf with the most
common value of
y
in
S
as a label.
ELSE
∀
A
find
a
that obtain the best
SplitCriterion
(
a
i
,S
)
.
Label
t
with
a
FOR each outcome
v
i
of
a
:
Set
Subtree
i
= TreeGrowing
(
σ
a
=
v
i
S, A, y
)
.
Connect the root node of
t
T
to
Subtree
i
with
an edge that is labeled as
v
i
a
i
∈
END FOR
END IF
RETURN TreePruning (
S
,
T
,
y
)
TreePruning (
S
,
T
,
y
)
Where:
S
- Training Set
y
- Target Feature
T
- The tree to be pruned
DO
Select a node
t
in
T
such that pruning it
maximally improve some evaluation criteria
=
Ø THEN
T
=
pruned
(
T,t
)
UNTIL
t
=Ø
RETURN
T
IF
t
Fig. 3.1 Top-down algorithmic framework for decision trees induction.
Search WWH ::
Custom Search