def graph():
dt_data = tree.export_graphviz(tree(),
out_file=None,
feature_names=iris.feature_names,
class_names=iris.target_names,
filled=True,
rounded=True,
special_characters=True)
graph = graphviz.Source(dt_data)
graph.render('output')
def getSimEntDataset(f,words,task):
data = open(f,'r')
lines = data.readlines()
examples = []
for i in lines:
i=i.strip()
if(len(i) > 0):
i=i.split('\t')
if len(i) == 3:
if task == "sim":
e = (tree(i[0], words), tree(i[1], words), float(i[2]))
examples.append(e)
elif task == "ent":
e = (tree(i[0], words), tree(i[1], words), i[2])
examples.append(e)
else:
raise ValueError('Params.traintype not set correctly.')
else:
print(i)
return examples
def getSentimentDataset(f,words):
data = open(f,'r')
lines = data.readlines()
examples = []
for i in lines:
i=i.strip()
if(len(i) > 0):
i=i.split('\t')
if len(i) == 2:
e = (tree(i[0], words), i[1])
examples.append(e)
else:
print(i)
return examples
def main(argv):
if FLAGS.dataset == 'toy':
train_X, train_y, test_X, test_y, num_classes = get_toy_dataset()
elif FLAGS.dataset == 'mnist':
train_X, train_y, test_X, test_y, num_classes = get_mnist()
train_pred = None
if FLAGS.method == 'knn':
pred = knn(train_X, train_y, test_X)
elif FLAGS.method == 'svm':
train_pred, pred = svm(train_X, train_y, test_X)
elif FLAGS.method == 'tree':
pred = tree(train_X, train_y, test_X)
elif FLAGS.method == 'boosting':
pred = boosting(train_X, train_y, test_X)
elif FLAGS.method == 'nn':
train_pred, pred = nn(train_X, train_y, test_X, num_classes)
if train_pred is not None:
print('Train Accuracy: %f' % compute_accuracy(train_pred, train_y))
print('Accuracy: %f' % compute_accuracy(pred, test_y))
# ('knn', neighbors.KNeighborsClassifier()),
# #SVM: http://scikit-learn.org/stable/modules/svm.html
# ('svc', svm.SVC(probability=True)),
# #xgboost: http://xgboost.readthedocs.io/en/latest/model.html
# ('xgb', XGBClassifier())
# ]
# vote_hard = ensemble.VotingClassifier(estimators = vote_est, voting='hard')
# vote_hard_cv = model_selection.cross_validate(vote_hard, data1[data1_x_bin], data1[Target], cv = cv_split)
# vote_hard.fit(data1[data1_x_bin], data1[Target])
# # print("VOTING_CLASSIFIER Parameters: ", dtree.get_params())
# # print("VOTING_CLASSIFIER Training w/bin score mean: {:.2f}", format(vote_hard_cv['train_score'].mean()))
# print("HARD_VOTING_CLASSIFIER Test w/bin score mean: {:.2f}", format(vote_hard_cv['test_score'].mean()))
# vote_soft = ensemble.VotingClassifier(estimators = vote_est, voting='soft')
# vote_soft_cv = model_selection.cross_validate(vote_soft, data1[data1_x_bin], data1[Target], cv = cv_split)
# vote_soft.fit(data1[data1_x_bin], data1[Target])
# # print("VOTING_CLASSIFIER Parameters: ", dtree.get_params())
# # print("VOTING_CLASSIFIER Training w/bin score mean: {:.2f}", format(vote_hard_cv['train_score'].mean()))
# print("SOFT_VOTING_CLASSIFIER Test w/bin score mean: {:.2f}", format(vote_soft_cv['test_score'].mean()))
data_val['Survived'] = tree(data_val).astype(int)
submit = data_val[['PassengerId', 'Survived']]
submit.to_csv("submission2.csv", index=False)
print(submit.info())
import pandas as pd
data = pd.read_pickle('/home/hudson/Downloads/prostate.df')
data.head(2)
#cell 5
y = data.values[:, -1]
print y.shape, Counter(y.tolist())
x = data.values[:, :-1]
print x.shape
#cell 6
## Task 1
(You can use DecisionTree implementation from scikit-learn.)
Try decision tree on the above dataset. consider different values for the max depth of the tree ('max_depth') and min number of samples required to be a leaf node ('min-samples_leaf'). Conduct 10-fold cross-validation and:
- plot training error and testing error v.s. tree depth
- plot training error and testing error v.s. min. sample for leaf nodes
Error should be measured by percentage of misclassification (i.e., return 'normal' for 'tumor' and vice versa).
#cell 7
from sklearn import tree
from sklearn import metrics
from sklearn.cross_validation import KFold
n_folds = 10
#10 fold cross-validation
kf = KFold(len(y), n_folds=n_folds)
def predict(obj):
p = tree().predict([obj])
print('this is ', p)
classifier.predict(testing_set.drop('class', axis=1)))
dataframe = pd.read_csv("diabetes.csv")
classes = [x for x in list(dataframe['class'].unique())]
# split dataframe (67% / 33%)
training_set = dataframe.sample(frac=0.67)
testing_set = dataframe[~dataframe.isin(training_set).all(1)]
print(training_set)
print(testing_set)
nn3 = knn(training_set, 3)
nn5 = knn(training_set, 5)
nn11 = knn(training_set, 11)
tree = tree(training_set)
gnb = gnb(training_set)
functions = [nn3, nn5, nn11, tree, gnb]
classifiers = ["3NN", "5NN", "11NN", "tree", "naive_bayes"]
scores = [get_score(x, testing_set) for x in functions]
## confusion matrices
for classifier, function, score in zip(classifiers, functions, scores):
print("\n" + classifier + " classifier")
print("accuracy = " + str(round(score * 100, 2)) + "%")
print(get_confusion_matrix(function, testing_set))
## bar chart
plt.bar(classifiers, scores, align='center')
plt.ylabel("score")
df_test_tmp = df_test.replace("male", 0).replace("female", 1)
df_test_tmp = df_test_tmp.replace("C", 0).replace("Q", 1).replace("S", 2)
df_test_tmp["Age"].fillna(df_test_tmp["Age"].median(), inplace=True)
df_test_tmp["Fare"].fillna(df_test_tmp["Fare"].median(), inplace=True)
df_test_tmp["Embarked"].fillna(df_test_tmp["Embarked"].median(), inplace=True)
test_data = df_test_tmp.loc[:, [
"Pclass", "Sex", "Age", "SibSp", "Parch", "Fare", "Embarked"
]]
# In[76]:
max_acc = 0
max_depth = 0
max_model = None
for i in range(1, 100):
ret = tree(i, train_data, train_target, valid_data, valid_target)
if max_acc < ret[1]:
max_acc = ret[1]
max_depth = i
max_model = ret[0]
print(max_depth, ",", max_acc)
# In[79]:
predicted = max_model.predict(test_data)
with open("predict_result_data.csv", "w") as f:
writer = csv.writer(f, lineterminator='\n')
writer.writerow(["PassengerId", "Survived"])
for pid, survived in zip(df_test["PassengerId"], predicted):
writer.writerow([pid, survived])
def diseases(node, depth):
if tree_.feature[node] != _tree.TREE_UNDEFINED:
disease_name = feature_name[node]
threshold = tree_.threshold[node]
print("\n" + disease_name + "?\n")
answer = input()
# Asking the user if they have the symptoms displayed
if answer == 'Yes':
val = 1
else:
val = 0
if val <= threshold:
diseases(tree_.children_left[node], depth + 1)
else:
symptoms_present.append(disease_name)
diseases(tree_.children_right[node], depth + 1)
else:
present_disease = probable_disease(tree_.value[node])
print("Possible disease: " + present_disease)
red_column = data.columns
symptoms_given = red_column[
data.loc[present_disease].values[0].nonzero()]
print("\nPresent symtomp: " + str(list(symptoms_present)))
print("\nKnown symptomps of the disease: " +
str(list(symptoms_given)))
diseases(0, 1)
tree(classifier, column)
return f1_score(y_test, clf.predict(X_test), average='macro')
def less_X():
X = pd.merge(X.iloc[:, 0:25],
X.iloc[:, 1024:1049],
how='outer',
left_index=True,
right_index=True)
X, y = shuffle(X, y, random_state=0)
X_train, X_valid, X_test = X[:int((0.6 * len(X)))], X[int((
0.6 * len(X))):int((0.8 * len(X)))], X[int((0.8 * len(X))):]
y_train, y_valid, y_test = y[:int((0.6 * len(X)))], y[int((
0.6 * len(X))):int((0.8 * len(X)))], y[int((0.8 * len(X))):]
randomforest(100, 50)
if __name__ == "__main__":
svm()
randomforest(100, 100)
randomforest(100, 50)
NB()
less_X()
plot_data = []
for i in xrange(0, 100):
plot_data.append(tree())
plt.plot(plot_data)