def BuildTree(DF, Splitter, Parent):
flag = 0 ## To Store if Current Node is a Leaf
if (len(DF.keys()) == 1): ## No Attribute Left to Split on
flag = 1
if ((Parent != None)
and (Parent.depth == Max_Depth)): ## Maximum Depth Reached
flag = 2
if (CalcNode(DF,
Splitter)[-1] == 0): ## Current Node is Perfectly Classified
flag = 3
if (flag == 0):
Split_Attr = BestAttr(DF, Splitter)
Parent.split = Split_Attr
Attr_Val = np.unique(DF[Split_Attr])
for val in Attr_Val:
Decision_Node = Node(Split_Attr,
parent=Parent,
value=val,
split=None,
child={},
outcome=None)
Parent.child[val] = Decision_Node
Leaf, Outcome = BuildTree(
DF[DF[Split_Attr] == val].reset_index(drop=True), Splitter,
Decision_Node)
if (Leaf == None):
Decision_Node.outcome = Outcome
else:
Val, Count = np.unique(DF[Class], return_counts=True)
Outcome = Val[np.argmax(Count)]
return None, Outcome
return Parent, None
def BuildTree(Parent, Splitter, Max_Depth):
if not Parent.is_leaf:
for subtree in Parent.children:
BuildTree(subtree, Splitter, Max_Depth)
return Parent, None
flag = 0 ## To Store if Current Node is a Leaf
if (len(Parent.df.keys()) == 1): ## No Attribute Left to Split on
flag = 1
if (not (Parent.is_root)
and (Parent.depth == Max_Depth)): ## Maximum Depth Reached
flag = 2
if (CalcNode(Parent.df[Class].values,
Splitter)[-1] == 0): ## Current Node is Perfectly Classified
flag = 3
if (flag == 0):
Split_Attr = BestAttr(Parent.df, Splitter)
Parent.split = Split_Attr
Attr_Val, Inv_Index = np.unique(Parent.df[Split_Attr],
return_inverse=True)
Hash = np.zeros(shape=(len(Attr_Val), len(Parent.df)), dtype=bool)
for i in range(len(Parent.df)):
Hash[Inv_Index[i], i] = True
for i in range(len(Attr_Val)):
Decision_Node = Node(Split_Attr,
parent=Parent,
df=Parent.df[Hash[i]],
value=Attr_Val[i],
split=None,
child={},
outcome=None)
Parent.child[Attr_Val[i]] = Decision_Node
Leaf, Outcome = BuildTree(Decision_Node, Splitter, Max_Depth)
if (Leaf == None):
Decision_Node.outcome = Outcome
else:
Val, Count = np.unique(Parent.df[Class], return_counts=True)
Outcome = Val[np.argmax(Count)]
return None, Outcome
return Parent, None
