def pruning(Tree,x,y,t):
if not isinstance(Tree, dict):
error = 0
right = 0
for tag in y:
if tag == Tree:
right += 1
else:
error += 1
return Tree,right,error
nodefN = Tree.keys()[0]
nodef = int(nodefN[7:])
canpruning = True
total1 = 0
total2 = 0
totalerror = 0
newTree = {nodefN:{}}
if (t[nodef] == 1): #discrete
fV = [data[nodef] for data in x]
fVType = set(fV)
for Vtype in fVType:
[subSetx,subSety,subSett] = splitSet_discrete(x,y,t,nodef,Vtype)
try:
[pTree, right, error ]= pruning(Tree[nodefN][Vtype],subSetx, subSety, subSett)
newTree[nodefN][Vtype] = pTree
if not isinstance(pTree,dict):
if pTree == name1:
total1 += right
total2 += error
else:
total1 += error
total2 += right
totalerror += error
else:
canpruning = False
except:
totalerror += len(subSetx)
if total1 < total2:
total1 += len(subSetx)
else:
total2 += len(subSetx)
else: #numerical
minv = 0.0
maxv = 1.0
delta = (maxv - minv) / NforNumerical + 0.0001
for i in range(NforNumerical):
[subSetx,subSety,subSett] = splitSet_numerical(x,y,t,nodef,minv+delta*i,minv+delta*(i+1))
if (len(subSetx) > 0):
try:
[pTree, right, error ]= pruning(Tree[nodefN][i],subSetx, subSety, subSett)
newTree[nodefN][i] = pTree
if not isinstance(pTree,dict):
if pTree == name1:
total1 += right
total2 += error
else:
total1 += error
total2 += right
totalerror += error
else:
canpruning = False
except:
totalerror += len(subSetx)
if total1 < total2:
total1 += len(subSetx)
else:
total2 += len(subSetx)
if canpruning and totalerror > min(total1,total2):
#print name1,name2,totalerror,total1,total2
if total1 < total2:
#print Tree,totalerror,total1,total2,name1
return name2,total2,total1
else:
#print Tree,totalerror,total1,total2,name2
return name1,total1,total2
return newTree,len(x)-totalerror,totalerror
def pruning(Tree, x, y, t):
if not isinstance(Tree, dict):
error = 0
right = 0
for tag in y:
if tag == Tree:
right += 1
else:
error += 1
return Tree, right, error
nodefN = Tree.keys()[0]
nodef = int(nodefN[7:])
canpruning = True
total1 = 0
total2 = 0
totalerror = 0
newTree = {nodefN: {}}
if (t[nodef] == 1): #discrete
fV = [data[nodef] for data in x]
fVType = set(fV)
for Vtype in fVType:
[subSetx, subSety,
subSett] = splitSet_discrete(x, y, t, nodef, Vtype)
try:
[pTree, right, error] = pruning(Tree[nodefN][Vtype], subSetx,
subSety, subSett)
newTree[nodefN][Vtype] = pTree
if not isinstance(pTree, dict):
if pTree == name1:
total1 += right
total2 += error
else:
total1 += error
total2 += right
totalerror += error
else:
canpruning = False
except:
totalerror += len(subSetx)
if total1 < total2:
total1 += len(subSetx)
else:
total2 += len(subSetx)
else: #numerical
minv = 0.0
maxv = 1.0
delta = (maxv - minv) / NforNumerical + 0.0001
for i in range(NforNumerical):
[subSetx, subSety,
subSett] = splitSet_numerical(x, y, t, nodef, minv + delta * i,
minv + delta * (i + 1))
if (len(subSetx) > 0):
try:
[pTree, right, error] = pruning(Tree[nodefN][i], subSetx,
subSety, subSett)
newTree[nodefN][i] = pTree
if not isinstance(pTree, dict):
if pTree == name1:
total1 += right
total2 += error
else:
total1 += error
total2 += right
totalerror += error
else:
canpruning = False
except:
totalerror += len(subSetx)
if total1 < total2:
total1 += len(subSetx)
else:
total2 += len(subSetx)
if canpruning and totalerror > min(total1, total2):
#print name1,name2,totalerror,total1,total2
if total1 < total2:
#print Tree,totalerror,total1,total2,name1
return name2, total2, total1
else:
#print Tree,totalerror,total1,total2,name2
return name1, total1, total2
return newTree, len(x) - totalerror, totalerror
