def main(argv):
usage = "\tusage: --dataset=[file_name] --split=[1-100] --classifier=[algorithm_name] --target=[index] \
--ordered=[column:val0|val1|val2,column2:val0|val1],etc --ignore=[col1,col2,etc] \
--mapped=[1,2,3,etc] --bins=[col:num_bins,col:num_bins] --normalize"
# get args
opts, args = getopt.getopt(argv[1:], "h", ["help", "dataset=", "split=", "classifier=", "target=", "ordered=",
"ignore=", "mapped=", "normalize", "bins="])
# handle args
data_set = class_name = "none"
split = 70
target = -1
ordered = {}
ignore = []
mapped = []
bins = {}
norm = False
for key, val in opts:
if key in ("-h", "--help"):
print(usage)
sys.exit()
elif key == "--dataset":
data_set = val
elif key == "--split":
split = int(val)
elif key == "--classifier":
class_name = val
elif key == "--target":
target = int(val)
elif key == "--ignore":
ignore = val.split(",")
for idx, s in enumerate(ignore):
ignore[idx] = int(s)
elif key == "--ordered":
entries = val.split(",")
for entry in entries:
entry_split = entry.split(":")
ordered[int(entry_split[0])] = entry_split[1].split("|")
elif key == "--mapped":
entries = val.split(",")
for entry in entries:
mapped.append(int(entry))
elif key == "--bins":
entries = val.split(",")
for entry in entries:
entry_split = entry.split(":")
bins[int(entry_split[0])] = int(entry_split[1])
elif key == "--normalize":
norm = True
else:
assert False, "unhandled option"
# Load the data set
data = None
# default load iris data
if data_set == "none":
print("No data set specified")
sys.exit()
# load from a csv file
else:
data = preprocessor.DataSet(data_set, ordered=ordered, target=target, split=split, ignore=ignore,
mapped=mapped, norm=norm, bins=bins)
# select the network to use
net = None
predict_targets = None
if class_name == "HardCoded":
net = classifier.HardCoded()
# train
net.train(data.train_attributes, data.train_targets, data.mapped_columns)
# predict
predictions = net.predict(data.test_attributes)
predict_targets = data.test_targets
elif class_name == "KNearestNeighbors":
net = classifier.KNearestNeighbor(k=1)
# train
net.train(data.train_attributes, data.train_targets, data.mapped_columns)
# predict
predictions = net.predict(data.test_attributes)
predict_targets = data.test_targets
elif class_name == "KNearestNeighbors_alt":
net = KNeighborsClassifier(n_neighbors=5)
# train
net.fit(data.train_attributes, data.train_targets.ravel())
# predict
predictions = net.predict(data.test_attributes)
predict_targets = data.test_targets
elif class_name == "DecisionTree":
net = classifier.ID3()
# train
net.train(data.train_attributes, data.train_targets)
net.output_tree(net.root, 0)
# predict
predictions = net.predict(data.test_attributes)
predict_targets = data.test_targets
elif class_name == "DecisionTree_alt":
net = tree.DecisionTreeClassifier()
# train
net.fit(data.train_attributes, data.train_targets.ravel())
# predict
predictions = net.predict(data.test_attributes)
predict_targets = data.test_targets
elif class_name == "Perceptron":
net = classifier.Perceptron(.1, [4, 4])
# train
net.train(data.train_attributes, data.train_targets)
# predict
predictions = net.predict(data.test_attributes)
predict_targets = data.test_targets
else:
print("Unrecognized classifier")
sys.exit(1)
# test the predictions
i = 0
num_right = 0
while i < len(predictions):
if predictions[i] == predict_targets[i]:
num_right += 1
i += 1
print("The number of correct predictions is: ", str(num_right / len(predictions) * 100), "%")
