def main(argv):
# Load a dataset containing many instances each with a set of attributes and a target value.
if len(argv) >= 2:
fileOrURL = argv[1]
isFile = False if fileOrURL.startswith('http') else True
if not isFile:
# Lets open the data from the url that has the data in a csv file format
# fileOrURL = "http://archive.ics.uci.edu/ml/machine-learning-databases/iris/bezdekIris.data"
f = io.StringIO(urllib.request.urlopen(fileOrURL).read().decode('utf-8')[:-1])
csv = numpy.genfromtxt(fileOrURL, delimiter=",", dtype=str)
numcols = len(csv[0])
# print ("CSV: ", csv)
data = csv[:, :-1]
targets = csv[:,-1]
else:
# Please use the popular Iris dataset (natively in scikit-learn).
iris = datasets.load_iris()
data = iris.data
targets = iris.target
# Randomize the order of the instances in the dataset. Don't forget that you need to keep the targets matched up with the approprite instance.
# This permutation will be used for both the data and the target so it will line up correctly.
perm = numpy.random.permutation(len(data))
data = data[perm]
targets = targets[perm]
# Split the data into two sets: a training set (70%) and a testing set (30%)
# Index of where to split (we want this to be an integer)
index = int(round(perm.size*.3))
test = perm[:index]
train = perm[index:]
# Instantiate your new classifier
# classifier = HCClassifier()
# classifier = KNNClassifier(3)
classifier = TreeClassifier()
# "Train" it with data
print ("Last Parameter: ", argv[-1])
if argv[-1] is not 'True':
# Before we make the tree we should make sure the data is simple (only a few bins per column)
# print ("Data before simplify: ", data)
data = csvSimplifier.simplify(data)
# print ("Data after simplify: ", data)
classifier.train(data[train], targets[train])
# Make "Predictions" on the test data
predictions = classifier.predict(data[test])
# Reset correct answers
correct = 0
# Count the answers that we got right
for (prediction, actual) in zip(predictions, test):
# print ("Prediction: ", prediction)
# print ("Actual: ", targets[actual])
if prediction == targets[actual]:
correct += 1
#Determine the accuracy of your classifier's predictions (reported as percentage)
print (correct/test.size*100)
# Create new public repository at GitHub and publish code
# Github repository: http://github.com/DanielRMiller/cs450tree
return
csv = numpy.genfromtxt(csvFile, delimiter=",", dtype=str)
numcols = len(csv[0])
data = csv[:, :-1]
targets = csv[:,-1]
# Randomize the order of the instances in the dataset. Don't forget that you need to keep the targets matched up with the approprite instance.
# This permutation will be used for both the data and the target so it will line up correctly.
perm = numpy.random.permutation(len(data))
data = data[perm]
targets = targets[perm]
# Split the data into two sets: a training set (70%) and a testing set (30%)
# Index of where to split (we want this to be an integer)
index = int(round(perm.size*.3))
test = perm[:index]
train = perm[index:]
# Instantiate your new classifier
classifier = TreeClassifier()
# "Train" it with data
# print ("Test Data: ", data[train])
classifier.train(data[train], targets[train], True)
# print ("Root: ", classifier.root.attribute)
# Make "Predictions" on the test data
predictions = classifier.predict(data[test])
# Reset correct answers
correct = 0
# Count the answers that we got right
for (prediction, actual) in zip(predictions, test):
if prediction == targets[actual]:
correct += 1
#Determine the accuracy of your classifier's predictions (reported as percentage)
print (correct/test.size*100)
