def Main (filename, ParentNode, MapParentNode) :
# repeat for all categories
for category in MapParentNode :
#check that the category is one class
if OneClass(filename, ParentNode, category)==False:
#compute the entropy
categoryEntropy=CategoryEntropy(filename,ParentNode.name,category)
# for all attributes not already in the tree
gains= dict()
for attribute in attributes:
if isParent(ParentNode,attribute)== False :
tmpMap={}
for key in (attributes[attribute]):
tmpMap[key]=[0,0]
# for each row of the file
with open(filename, 'rb') as csvfile:
csvreader = csv.reader(csvfile, delimiter=',')
for row in csvreader:
if row[ParentNode.name]== category :
for key in tmpMap :
if row[attribute] == key :
if row[14] == ">50K":
tmpMap[key][0]=tmpMap[key][0]+1
else :
tmpMap[key][1]=tmpMap[key][1]+1
# Gain
Entropy=0
for key in attributes[attribute] :
nb=tmpMap[key][0]+tmpMap[key][1]
if tmpMap[key][0] !=0 and tmpMap[key][1] !=0:
Entropy=Entropy+((tmpMap[key][0]/float(nb))*math.log(tmpMap[key][0]/float(nb),2) + (tmpMap[key][1]/float(nb))*math.log(tmpMap[key][1]/float(nb),2))*float(nb)/len(attributes[ParentNode.name][category])
else :
Entropy =0
gain= -categoryEntropy + Entropy
gains[attribute]=gain
if(len(gains)!=0) :
# obtain the highest gain
att=-1
max_val = max(gains.itervalues())
for k, v in gains.iteritems() :
if v == max_val :
att=k
# Create the Node
NewNode= Node()
NewNode.name= att
# add parents
NewNode.parents= ParentNode.parents.copy()
NewNode.parents[ParentNode.name]= category
NewNode.parents[NewNode.name]='None'
# Add the node as a child of parent Node
ParentNode.children[category]=NewNode
# repeat for children
Main(filename,NewNode,attributes.get(NewNode.name))
else :
#No more attribute, but still not a class
NewNode = Node()
NewNode.name= -1
NewNode.parents= ParentNode.parents.copy()
NewNode.parents[ParentNode.name]= category
NewNode.parents[NewNode.name]='None'
# Add the node as a child of parent Node
ParentNode.children[category]=NewNode
else :
# the category is a class
NewNode = Node()
NewNode.name= -1
bool,NewNode.Class=OneClass(filename, ParentNode, category)
NewNode.parents= ParentNode.parents.copy()
NewNode.parents[ParentNode.name]= category
NewNode.parents[NewNode.name]='None'
# Add the node as a child of parent Node
ParentNode.children[category]=NewNode
gain.append(0)
gain.append(Gain("adult1.data", education,3,nbRows("adult1.data"), EntropyAllData("adult1.data",nbRows("adult1.data"))))
# ignore Education Number attribute
gain.append(0)
gain.append(Gain("adult1.data", maritalStatus,5,nbRows("adult1.data"), EntropyAllData("adult1.data",nbRows("adult1.data"))))
gain.append(Gain("adult1.data", occupation,6,nbRows("adult1.data"), EntropyAllData("adult1.data",nbRows("adult1.data"))))
gain.append(Gain("adult1.data", relationship,7,nbRows("adult1.data"), EntropyAllData("adult1.data",nbRows("adult1.data"))))
gain.append(Gain("adult1.data", race,8,nbRows("adult1.data"), EntropyAllData("adult1.data",nbRows("adult1.data"))))
gain.append(Gain("adult1.data", sex,9,nbRows("adult1.data"), EntropyAllData("adult1.data",nbRows("adult1.data"))))
gain.append(Gain("adult1.data", capitalGain,10,nbRows("adult1.data"), EntropyAllData("adult1.data",nbRows("adult1.data"))))
gain.append(Gain("adult1.data", capitalLoss,11,nbRows("adult1.data"), EntropyAllData("adult1.data",nbRows("adult1.data"))))
gain.append(Gain("adult1.data", hours,12,nbRows("adult1.data"), EntropyAllData("adult1.data",nbRows("adult1.data"))))
gain.append(Gain("adult1.data", nativeCountry,13,nbRows("adult1.data"), EntropyAllData("adult1.data",nbRows("adult1.data"))))
# Initialisation
Tree=Node()
Tree.name=gain.index(max(gain))
Tree.parents[Tree.name]='None'
# Build decision Tree
Main ("adult1.data",Tree,attributes[Tree.name])
#Apply Decision Tree
DecisionTree("Test", "Test1", Tree)
#output
output( "Test1","Test2")
