def test(dataset):
prunedTest = []
prunedTrain = []
testAcc = []
trainSetAcc = []
data = processDataset(dataset)
# random pick data form dataset and test, return the mean accuracy
for i in range(35):
random.shuffle(data)
trainSet = data[:2 * len(data) / 3]
testSet = data[2 * len(data) / 3:]
tree = decisionTree.buildTree(trainSet, 'Class')
accTrain = decisionTree.test(tree, trainSet)
accTest = decisionTree.test(tree, testSet)
prunedTree = decisionTree.pruneTree(tree, testSet)
accPruneTestset = decisionTree.test(prunedTree, testSet)
accPruneTrainset = decisionTree.test(prunedTree, trainSet)
trainSetAcc.append(accTrain)
testAcc.append(accTest)
prunedTrain.append(accPruneTrainset)
prunedTest.append(accPruneTestset)
print
print "Before the pruning,the trainSet accuracy : ", sum(trainSetAcc) / len(trainSetAcc)
print "Before the pruning,the testSet accuracy : ", sum(testAcc) / len(testAcc)
print "After the pruning, the trainSet accuracy : ", sum(prunedTrain) / len(prunedTrain)
print "After the pruning, the testSet accuracy : ", sum(prunedTest) / len(prunedTest)
def __main_(): #todo to make this real, add an extra _ at the end of the name
l = sys.argv[0]
k = sys.argv[1]
trainingSet = sys.argv[2]
validationSet = sys.argv[3]
testSet = sys.argv[4]
toPrint = sys.argv[5]
tree = dt.buildTree(trainingSet, "e")
print("Test results, entropy: " + tree.test(testSet))
print("Test results, post pruning, entropy" +
dt.postPruning(tree, l, k, validationSet))
if toPrint == "yes":
tree.printTree()
tree = dt.buildTree(trainingSet, "v")
print("Test results, VI: " + tree.test(testSet))
print("Test results, post pruning, VI" +
dt.postPruning(tree, l, k, validationSet))
if toPrint == "yes":
tree.printTree()
import decisionTree
import sys
trainIns, trainOuts = decisionTree.loadData(sys.argv[1])
testIns, testOuts = decisionTree.loadData(sys.argv[2])
tree = decisionTree.buildTree(trainIns, trainOuts, 1)
(theta, feature, gain, leftCategory, rightCategory) = tree
print("Learned Tree: ")
print("\tFeature " + str(feature) + ": (Information gain: " + str(gain) + ")")
print("\t| > " + str(theta) + " -> " + str(leftCategory))
print("\t| < " + str(theta) + " -> " + str(rightCategory))
trainErrs = decisionTree.runTree(tree, trainIns, trainOuts)
testErrs = decisionTree.runTree(tree, testIns, testOuts)
print("\nTraining Error:\t" + str(trainErrs * 100. / len(trainOuts)) + "%")
print("Testing Error:\t" + str(testErrs * 100. / len(testOuts)) + "%")
inp = list(i[1].iloc[j][:-1])
expect = i[1].iloc[j][-1:]
nn.setInput(inp)
nn.allProcess()
o = nn.classify()
if o == int(expect):
success += 1
# nn.getLayerList()
t_e = success / float(len(i[1]))
print("This error :", t_e)
error.append(t_e)
else: #Decision tree
input.normalize(i[0])
input.normalize(i[1])
target = [xp for xp in range(21)]
root = dt.DecisionTree.Node(target)
root.avaiable_attr = list(i[0])[:-1]
dt.buildTree(i[0], root, 'G3')
success = 0
for j in range(len(i[1])):
result = root.classify(i[1].iloc[j])
if result == int(i[1].iloc[j]['G3']):
success += 1
t_e = success / float(len(i[1]))
print("This error :", t_e)
error.append(t_e)
a_error = sum(error) / 10.0
print('ERROR is', a_error)
import pandas as pd
import scipy as sc
import numpy as np
import math
import decisionTree as dt
import sys
tree = dt.buildTree("training_set_1.csv", "e")
tree.printTree()
tree = dt.buildTree("training_set_1.csv", "v")
tree.printTree()
print(tree.test("test_set_1.csv"))
#dt.buildTree("training_set_1.csv", "v")
def findEntropy(matches, mismatches):
total = matches + mismatches
print(total)
first = -1 * (matches / total) * np.log2(matches / total)
print(first)
second = -1 * (mismatches / total) * np.log2(mismatches / total)
print(second)
return first + second
# n = dt.node("asdf")
# n.total = 4
# n.mismatchesOne = 4
# n.matchesOne = 0
# n.mismatchesZero = 0
# n.matchesZero = 0