def classification_linear_svm(tweets,
train_index,
test_index,
labels_train,
random_state=None):
"""Classifies using SVM as classifier
"""
#Representation
tfidf_parser = TfidfVectorizer(tokenizer=tokenize,
lowercase=False,
analyzer='word')
tweets_train = [tweets[tweet_index] for tweet_index in train_index]
tweets_test = [tweets[tweet_index] for tweet_index in test_index]
train_sparse_matrix_features_tfidf = tfidf_parser.fit_transform(
tweets_train)
test_sparse_matrix_features_tfidf = tfidf_parser.transform(tweets_test)
classifier = LinearSVC(multi_class="ovr", random_state=random_state)
print("Start SVM training")
classifier = classifier.fit(train_sparse_matrix_features_tfidf,
labels_train)
print("Finish SVM training")
y_labels = classifier.predict(test_sparse_matrix_features_tfidf)
return y_labels
class Expander_LDA_multiclass(Expander_LDA_cossim):
"""
take LDA vectors of labelled articles and do a multi-class
classification for deciding where the LDA of the test text belongs
"""
def __init__(self, ldaModelAll, expander_type=AcronymExpanderEnum.LDA_multiclass):
Expander_LDA_cossim.__init__(self, ldaModelAll, expander_type)
self.classifier = LinearSVC()
def transform(self, X):
results = Expander_LDA_cossim.transform(self, X)
return [self._getDenseVector(item) for item in results]
def _getDenseVector(self, sparse_vec):
return sparse2full(sparse_vec, self.ldaModel.num_topics)
def fit(self, X_train, y_train):
self.classifier.fit(X_train, y_train)
def predict(self, X_test, acronym):
labels = self.classifier.predict(X_test)
decisions = self.classifier.decision_function(X_test)
confidences = self._getConfidencesFromDecisionFunction(labels, decisions)
return labels, confidences
def fit(self, data, args):
self.model = LinearSVC()
with Timer() as t:
self.model.fit(data.X_train, data.y_train)
return t.interval
def applyModelandfit(tweet_list, tweet_label_list,all_tweet,model_name,filename):
X, y, tweet_id_list = buildMatrixTrainAndTest(tweet_list, tweet_label_list, all_tweet)
X_train = X[:len(y),:]
y_train = y
X_test = X[len(y):,:]
tweet_id_list_test = tweet_id_list[len(y):]
print "number of training tweets are ", X_train.shape, len(y_train)
if model_name == 'SVM':
clf = LinearSVC(penalty="l1", dual=False, tol=1e-7)
clf.fit(X_train, y_train)
elif model_name == 'NaiveBayes':
clf = GaussianNB()
clf.fit(X_train.todense(), y_train)
elif model_name == 'LogisticRegression':
clf = LogisticRegression(C=1.0, penalty='l1', tol=0.01)
clf.fit(X_train.toarray(), y_train)
else:
raise Exception("The model name is incorrect!!!")
y_pred = clf.predict(X_test)
print 'length of predict data is ', len(y_pred)
with open(RESULT_FOLDER+'/'+filename+'_c.csv','wb') as fp:
writer = csv.writer(fp, delimiter =",",quoting=csv.QUOTE_MINIMAL)
for i, tweetid in enumerate(tweet_id_list_test):
writer.writerow([tweetid, all_tweet[tweetid], y_pred[i]])
class LinearSVCImpl():
def __init__(self, penalty='l2', loss='squared_hinge', dual=True, tol=0.0001, C=1.0, multi_class='ovr', fit_intercept=True, intercept_scaling=1, class_weight='balanced', verbose=0, random_state=None, max_iter=1000):
self._hyperparams = {
'penalty': penalty,
'loss': loss,
'dual': dual,
'tol': tol,
'C': C,
'multi_class': multi_class,
'fit_intercept': fit_intercept,
'intercept_scaling': intercept_scaling,
'class_weight': class_weight,
'verbose': verbose,
'random_state': random_state,
'max_iter': max_iter}
def fit(self, X, y=None):
self._sklearn_model = SKLModel(**self._hyperparams)
if (y is not None):
self._sklearn_model.fit(X, y)
else:
self._sklearn_model.fit(X)
return self
def predict(self, X):
return self._sklearn_model.predict(X)
def get_svm_score(w, b_h, dataset):
"""
Given a trained RBM, get the classification score of a linear SVM trained on the hidden Representation
:param w: Weights
:param b_h: Hidden biases
:param dataset: A Dataset object
:return: A scalar score
"""
proj_training_data = sigm(dataset.training_set.input.dot(w)+b_h)
classifier = LinearSVC()
classifier.fit(proj_training_data, dataset.training_set.target)
proj_test_data = sigm(dataset.test_set.input.dot(w)+b_h)
predicted_labels = classifier.predict(proj_test_data)
score = percent_correct(dataset.test_set.target, predicted_labels)
return score
def get_svm_score(w, b_h, dataset):
"""
Given a trained RBM, get the classification score of a linear SVM trained on the hidden Representation
:param w: Weights
:param b_h: Hidden biases
:param dataset: A Dataset object
:return: A scalar score
"""
proj_training_data = sigm(dataset.training_set.input.dot(w)+b_h)
classifier = LinearSVC()
classifier.fit(proj_training_data, dataset.training_set.target)
proj_test_data = sigm(dataset.test_set.input.dot(w)+b_h)
predicted_labels = classifier.predict(proj_test_data)
score = percent_correct(dataset.test_set.target, predicted_labels)
return score
class CreateLinearSVC(CreateModel):
def fit(self, data, args):
self.model = LinearSVC()
with Timer() as t:
self.model.fit(data.X_train, data.y_train)
return t.interval
def predict(self, data):
assert self.model is not None
with Timer() as t:
self.predictions = self.test(data)
return t.interval
def fit(self, X, y=None):
self._sklearn_model = SKLModel(**self._hyperparams)
if (y is not None):
self._sklearn_model.fit(X, y)
else:
self._sklearn_model.fit(X)
return self
def test_model_within_optimizer(self):
pipe = Pipeline([('reduce_dim', PCA()), ('classify', LinearSVC())])
n_features_options = [2, 4, 8]
c_options = [1, 10, 100, 1000]
param_grid = [
{
'reduce_dim': [PCA(iterated_power=7),
NMF()],
'reduce_dim__n_components': n_features_options,
'classify__C': c_options
},
{
'reduce_dim': [SelectKBest(chi2)],
'reduce_dim__k': n_features_options,
'classify__C': c_options
},
]
grid = GridSearchCV(pipe, cv=3, n_jobs=1, param_grid=param_grid)
digits = load_digits()
grid.fit(digits.data, digits.target)
try:
Porter(grid, language='java')
except ValueError:
self.assertTrue(False)
else:
self.assertTrue(True)
def train_with_svm(self):
rbm = BernoulliRBM(random_state=0, verbose=False)
svc = LinearSVC(C=10.0,class_weight='balanced',max_iter=100)
classifier = Pipeline(steps=[('rbm', rbm), ('svm', svc)])
rbm.learning_rate = 0.05
rbm.n_iter = 30
# More components tend to give better prediction performance, but larger
# fitting time
rbm.n_components = 100
classifier.fit(self.X, self.Y)
self.classifier = classifier
joblib.dump(classifier,"rbm.pkl")
def runandsaveModel(tweet_list, tweet_label_list,model_name):
X, y, vectorizer= buildMatrixTrain(tweet_list, tweet_label_list)
print "number of training tweets are ", X.shape, len(y)
#trainning the model
if model_name == 'SVM':
clf = LinearSVC(penalty="l1", dual=False, tol=1e-7)
clf.fit(X, y)
elif model_name == 'NaiveBayes':
clf = GaussianNB()
clf.fit(X.todense(), y)
elif model_name == 'LogisticRegression':
clf = LogisticRegression(C=1.0, penalty='l1', tol=0.01)
clf.fit(X.toarray(), y)
else:
raise Exception("The model name is incorrect!!!")
#save the model
model = Model(model_name, clf, vectorizer)
with open(RESULT_FOLDER+"/"+model_name+"_model.m","wb") as pf:
pickle.dump(model,pf)
print model_name, "is saved at", RESULT_FOLDER+"/"+model_name+"_model.m"
def __init__(self, penalty='l2', loss='squared_hinge', dual=True, tol=0.0001, C=1.0, multi_class='ovr', fit_intercept=True, intercept_scaling=1, class_weight='balanced', verbose=0, random_state=None, max_iter=1000):
self._hyperparams = {
'penalty': penalty,
'loss': loss,
'dual': dual,
'tol': tol,
'C': C,
'multi_class': multi_class,
'fit_intercept': fit_intercept,
'intercept_scaling': intercept_scaling,
'class_weight': class_weight,
'verbose': verbose,
'random_state': random_state,
'max_iter': max_iter}
self._wrapped_model = SKLModel(**self._hyperparams)
def single_model_tuning(modelname, fold_nr):
"""
The thread function that can be used for finding the best model hyperparameters, for a single, non-ensemble model,
for a fixed preprocessor, this method requires the data to be split in folds first.
parameters:
:param str modelname: The name of the model to test.
:param int fold_nr: The number of the fold.
:return list<dict> results: A list of dictionaries containing the parameter setting and the mae.
"""
# Init a best mae so far (for printing purposes)
best = 10
try:
log('Fold: ' + str(fold_nr) + ': Loaded the cached preprocessed data.')
X_train, X_val, y_train, y_val, rev_val = load_fold(fold_nr)
except IOError:
log('Fold: ' + str(fold_nr) + 'run "python kfold_prepr.py" first')
results = []
# Tune a model based on the command line argument
if modelname == 'log':
par = ParameterGrid({
'logistic__C': np.logspace(-5.0, 5.0, num=11),
'logistic__tol': np.logspace(-5.0, 5.0, num=11)
})
for a in list(par):
logistic = LogisticRegression(solver='sag',
n_jobs=NUM_THEADS,
C=a['logistic__C'],
tol=a['logistic__tol'])
logistic.fit(X_train, y_train)
predictions_val = logistic.predict(X_val)
mae = mean_absolute_error(predictions_val, y_val)
results.append({
'logistic__C': a['logistic__C'],
'logistic__tol': a['logistic__tol'],
'mae': mae
})
elif modelname == 'ridge':
par = ParameterGrid({'ridge__alpha': np.logspace(-5.0, 5.0, num=11)})
for a in list(par):
ridge = OrdinalRidge(a['ridge__alpha'])
ridge.fit(X_train, y_train)
predictions_val = ridge.predict(X_val)
mae = mean_absolute_error(predictions_val, y_val)
results.append({'ridge__alpha': a['ridge__alpha'], 'mae': mae})
elif modelname == 'svc':
par = ParameterGrid({
'svc__C': np.logspace(-5.0, 5.0, num=11),
'svc__tol': np.logspace(-5.0, 5.0, num=11)
})
for a in list(par):
svc = LinearSVC(C=a['svc__C'], tol=a['svc__tol'])
svc.fit(X_train, y_train)
predictions_val = svc.predict(X_val)
mae = mean_absolute_error(predictions_val, y_val)
results.append({
'svc__C': a['svc__C'],
'svc__tol': a['svc__tol'],
'mae': mae
})
elif modelname == 'lad':
par = ParameterGrid({
'lad__C': np.logspace(-5.0, 5.0, num=11),
'lad__tol': np.logspace(-5.0, 5.0, num=11)
})
for a in list(par):
svr_ = svm.LinearSVR(loss='squared_epsilon_insensitive')
svr = LAD(svr_) # use mord for rounding and clipping
svr.fit(X_train, y_train)
predictions_val = svr.predict(X_val)
mae = mean_absolute_error(predictions_val, y_val)
results.append({
'lad__C': a['lad__C'],
'lad__tol': a['lad__tol'],
'mae': mae
})
elif modelname == 'final':
# This is the tuning of the final ensemble, with fixing 0 rating predictions
par = ParameterGrid({
'logistic_lbfgs__C':
np.logspace(-5.0, 5.0, num=11),
'logistic_lbfgs__tol':
np.logspace(-5.0, 5.0, num=11),
'logistic_lbfgs_multinom__C':
np.logspace(-5.0, 5.0, num=11),
'logistic_lbfgs_multinom__tol':
np.logspace(-5.0, 5.0, num=11),
'logistic_sag_balanced__C':
np.logspace(-5.0, 5.0, num=11),
'logistic_sag_balanced__tol':
np.logspace(-5.0, 5.0, num=11)
})
ensemble = VotingClassifier(estimators=[
('logistic_lbfgs',
LogisticRegression(solver='lbfgs',
n_jobs=NUM_THEADS,
C=5,
tol=0.01)),
('logistic_lbfgs_multinom',
LogisticRegression(solver='lbfgs',
n_jobs=NUM_THEADS,
C=5,
tol=0.01,
multi_class='multinomial')),
('logistic_sag_balanced',
LogisticRegression(solver='sag',
n_jobs=NUM_THEADS,
C=5,
tol=0.01,
class_weight='balanced')),
],
voting='soft',
weights=[1, 1, 1])
for a in list(par):
ensemble.set_params(**a)
ensemble.fit(X_train, y_train)
predictions_val = ensemble.predict(X_val)
predictions_val = fix_zero_predictions(predictions_val, rev_val)
mae = mean_absolute_error(predictions_val, y_val)
temp = a
temp['mae'] = mae
if mae < best:
print temp
best = mae
results.append(temp)
elif modelname == 'lbfgs_bal':
clf = LogisticRegression(solver='lbfgs',
n_jobs=NUM_THEADS,
class_weight='balanced')
par = ParameterGrid({
'C': np.logspace(-1.0, 1.0, num=5),
'tol': np.logspace(-3.0, -1.0, num=3)
})
for a in list(par):
clf.set_params(**a)
clf.fit(X_train, y_train)
predictions_val = clf.predict(X_val)
predictions_val = fix_zero_predictions(predictions_val, rev_val)
mae = mean_absolute_error(predictions_val, y_val)
temp = a
temp['mae'] = mae
if mae < best:
print temp
best = mae
results.append(temp)
elif modelname == 'lbfgs_multi':
clf = LogisticRegression(solver='lbfgs',
n_jobs=NUM_THEADS,
multi_class='multinomial')
par = ParameterGrid({
'C': np.logspace(-5.0, 5.0, num=11),
'tol': np.logspace(-5.0, 5.0, num=11)
})
for a in list(par):
clf.set_params(**a)
clf.fit(X_train, y_train)
predictions_val = clf.predict(X_val)
predictions_val = fix_zero_predictions(predictions_val, rev_val)
mae = mean_absolute_error(predictions_val, y_val)
temp = a
temp['mae'] = mae
if mae < best:
print temp
best = mae
results.append(temp)
elif modelname == 'sag_bal':
clf = LogisticRegression(solver='sag',
n_jobs=NUM_THEADS,
class_weight='balanced')
par = ParameterGrid({
'C': np.logspace(-5.0, 5.0, num=11),
'tol': np.logspace(-5.0, 5.0, num=11)
})
for a in list(par):
clf.set_params(**a)
clf.fit(X_train, y_train)
predictions_val = clf.predict(X_val)
predictions_val = fix_zero_predictions(predictions_val, rev_val)
mae = mean_absolute_error(predictions_val, y_val)
temp = a
temp['mae'] = mae
if mae < best:
print temp
best = mae
results.append(temp)
elif modelname == 'nb':
clf = MultinomialNB()
par = ParameterGrid(
{'alpha': [0, 0.01, 0.05, 0.1, 0.2, 0.3, 0.4, 0.5, 1, 1.5]})
for a in list(par):
clf.set_params(**a)
clf.fit(X_train, y_train)
predictions_val = clf.predict(X_val)
predictions_val = fix_zero_predictions(predictions_val, rev_val)
mae = mean_absolute_error(predictions_val, y_val)
temp = a
temp['mae'] = mae
if mae < best:
print temp
best = mae
results.append(temp)
else:
print "model name not defined"
return None
return results
def main_preprop():
"""
The main method that can be used for finding the best preprocessor hyperparameters,
for a fixed model.
"""
# Initialize the result array
results = []
# Set the search range in a parameter grid, all possible combinations will be tested
params_preprocessor = {
'a_value': range(10, 15),
'epsilon': [0.1, 0.01, 0.001],
'reduction_level': [0.01, 0.02, 0.05, 0.1, 0.15, 0.2, 0.3, 0.5]
}
# Create a preprocessor object (tweaked for tuning hyperparameters)
preprocessor_object = Preprocessor2()
# Define the models
logistic = LogisticRegression(solver='sag',
n_jobs=NUM_THEADS,
C=5,
tol=0.01)
ridge = OrdinalRidge(10)
svc = LinearSVC(C=0.5)
svr = svm.LinearSVR(loss='squared_epsilon_insensitive')
lad = LAD(svr)
ensemble = VotingClassifier(estimators=[('logistic', logistic),
('ridge', ridge), ('svc', svc),
('lad', lad)],
voting='hard',
weights=[1, 1, 1, 1])
# load the data
train_data, test_data = preprocessor_object.load()
# Convert the data to tf-idf matrices
X_train_full, X_val_full, y_train, y_val = preprocessor_object.fit_data(
train_data)
# Test the preprocessor for all possible hyperparameter combinations
for x in params_preprocessor['a_value']:
for y in params_preprocessor['epsilon']:
sigma_values = preprocessor_object.compute_sigma_values(
X_train_full, y_train, x, y)
for z in params_preprocessor['reduction_level']:
X_train, X_val, X_test = preprocessor_object.remove_features(
sigma_values, z, X_train_full, X_val_full, test_data)
ensemble.fit(X_train, y_train)
predictions_val = ensemble.predict(X_val)
mae = mean_absolute_error(predictions_val, y_val)
print strftime("%H:%M:%S") + ' -> a: ' + str(
x) + '; epsilon: ' + str(y) + '; red lvl: ' + str(
z) + '; MAE: ' + str(mae)
results.append({
'a_value': x,
'epsilon': y,
'reduction_level': z,
'mae': mae
})
print results
log('That\'s all folks!')
def runModel(X, y, S_data, model_name):
f = open('r_' + "_" + model_name + '.txt', 'w')
auc_score_all = []
fold = S_data[1]
Index_gen = S_data[0]
label = ['0.0', '1.0']
# note that python does not copy the generator,
# so when it's in the end of the for loop the generator for S_data is also extruded!
print 'Running', model_name
for exp in range(0, fold):
print "=" * 80, "\n", "experiment =", exp
# getting one fold of indices from the index generator
train_index, test_index = Index_gen.next()
X_train, X_test = X.iloc[train_index, :], X.iloc[test_index, :]
y_train, y_test = y.iloc[train_index], y.iloc[test_index]
# fitting the model
# 1 fit a model
# 2 prediction
if model_name == 'SVM':
# LinearSVC take care of the multi class response by using one vs others method
clf = LinearSVC(random_state=0).fit(X_train, y_train)
y_pred = clf.predict(X_test)
elif model_name == 'NaiveBayes':
clf = GaussianNB()
clf.fit(X_train.to_dense(), y_train)
y_pred = clf.predict(X_test.to_dense())
elif model_name == 'LogisticRegression':
clf = LogisticRegression(C=1.0, penalty='l1', tol=0.01)
clf.fit(X_train.as_matrix(), y_train)
y_pred = clf.predict(X_test.as_matrix())
else:
raise Exception("The model name is incorrect!!!")
### 2.4 eval
auc_score = roc_auc_score(y_test, y_pred, average=None)
auc_score_all.append(auc_score)
auc_ave = mean(array(auc_score_all), 0)
print >> f, model_name, '\n', "=" * 80
print >> f, 'avg auc = ', auc_ave
#path = '/media/robbis/DATA/fmri/monks'
path = '/home/carlos/mount/megmri03/monks'
subjects = os.listdir(path)
subjects = [s for s in subjects if s.find('.') == -1 and s.find('_') == -1]
# Load monk data in the form of n_samples x n_voxels x n_time
ds, _, _ = load_subject_ds(
path,
subjects[:1],
#os.path.join(path, 'subjects.csv'),
'meditation_permut1.conf',
'fmri',
prepro=MonksPreprocessingPipeline(),
roi_labels=atlas_dict)
clf = make_pipeline(StandardScaler(), LinearSVC(C=1))
time_gen = GeneralizingEstimator(clf, scoring='accuracy', n_jobs=20)
ds = SampleSlicer({'group': ['E']}).transform(ds)
scores_dict = {}
# Generalization of time
for network in os.listdir(path_templates):
network = network[:-21]
ds_network = FeatureSlicer({network: ['!0']}).transform(ds)
n_samples, n_voxels = ds_network.shape
data = ds_network.samples.reshape(-1, 135, n_voxels)
X = np.rollaxis(data, 1, 3)
y = np.arange(data.shape[0]) % 2
def setUp(self):
super(LinearSVCJavaTest, self).setUp()
self.estimator = LinearSVC(C=1., random_state=0)
def __init__(self, ldaModelAll, expander_type=AcronymExpanderEnum.LDA_multiclass):
Expander_LDA_cossim.__init__(self, ldaModelAll, expander_type)
self.classifier = LinearSVC()
'KernelRidge':KernelRidge(), 'LSHForest':LSHForest(), 'LabelPropagation':LabelPropagation(), 'LabelSpreading':LabelSpreading(), 'Lars':Lars(), 'LarsCV':LarsCV(), 'Lasso':Lasso(), 'LassoCV':LassoCV(), 'LassoLars':LassoLars(), 'LassoLarsCV':LassoLarsCV(), 'LassoLarsIC':LassoLarsIC(), 'LatentDirichletAllocation':LatentDirichletAllocation(), 'LedoitWolf':LedoitWolf(), 'LinearDiscriminantAnalysis':LinearDiscriminantAnalysis(), 'LinearRegression':LinearRegression(), 'LinearSVC':LinearSVC(), 'LinearSVR':LinearSVR(), 'LocallyLinearEmbedding':LocallyLinearEmbedding(), 'LogisticRegression':LogisticRegression(), 'LogisticRegressionCV':LogisticRegressionCV(), 'MDS':MDS(), 'MLPClassifier':MLPClassifier(), 'MLPRegressor':MLPRegressor(), 'MaxAbsScaler':MaxAbsScaler(), 'MeanShift':MeanShift(), 'MinCovDet':MinCovDet(), 'MinMaxScaler':MinMaxScaler(), 'MiniBatchDictionaryLearning':MiniBatchDictionaryLearning(), 'MiniBatchKMeans':MiniBatchKMeans(), 'MiniBatchSparsePCA':MiniBatchSparsePCA(), 'MultiTaskElasticNet':MultiTaskElasticNet(),
def setUp(self):
super(LinearSVCTest, self).setUp()
mdl = LinearSVC(C=1., random_state=0)
self._port_model(mdl)
def runModel(X, y, model_name):
nFolders = 5
accs = []
precs = []
recalls = []
F1s = []
n = X.shape[0]
for exp in range(0, nFolders):
print '\n\n============================================================================================\nexperiment' , exp
### 2.1 split training and testing data
start = (int)((1-(exp+1) * 1.0/nFolders)*n)
end = (int)((1-exp * 1.0/nFolders)*n)
#print n, start, end
X_train, y_train, X_test, y_test = splitTrainTest(X, y, start, end)
print 'Running', model_name
if model_name == 'SVM':
### 2.2 build classifier
clf = LinearSVC(penalty="l1", dual=False, tol=1e-7)
clf.fit(X_train, y_train)
### 2.3 predict
y_pred = clf.predict(X_test)
elif model_name == 'NaiveBayes':
clf = GaussianNB()
clf.fit(X_train.todense(), y_train)
y_pred = clf.predict(X_test.todense())
elif model_name == 'LogisticRegression':
clf = LogisticRegression(C=1.0, penalty='l1', tol=0.01)
clf.fit(X_train.toarray(), y_train)
y_pred = clf.predict(X_test.toarray())
else:
raise Exception("The model name is incorrect!!!")
### 2.4 eval
acc, prec, recall, F1 = eval(y_test, y_pred)
print 'Acc = ', acc;
print 'Precision =', prec;
print 'Recall=', recall;
print 'F1 =', F1
accs.append(acc)
precs.append(prec)
recalls.append(recall)
F1s.append(F1)
print '\n\n\n'
print 'avg Acc = ', sum(accs)/len(accs)
print 'avg Precision = ', sum(precs)/len(precs)
print 'avg Recall = ', sum(recalls)/len(recalls)
print 'avg F1 = ', sum(F1s)/len(F1s)
return sum(accs)/len(accs), sum(precs)/len(precs), sum(recalls)/len(recalls), sum(F1s)/len(F1s)
n_classes = 2
FEAT_F33 = "F33"
ROLL_LEN = 10
PANTHEON_SIZE = 10
n_users = 1000
max_runs = None #10000
percTest = 0.1
predictors = [
# DummyClassifier(strategy="stratified"),
# DummyClassifier(strategy="uniform"),
# BernoulliNB(),
# SVC(kernel='rbf', max_iter=10000, class_weight="balanced", verbose=1),
LinearSVC(max_iter=50),
MLPClassifier(max_iter=50, nesterovs_momentum=True,
early_stopping=True), #, activation="logistic"),
# LogisticRegression(class_weight='balanced'),
RandomForestClassifier(class_weight="balanced"),
# ExtraTreeClassifier(),
# AdaBoostClassifier(),
# DecisionTreeClassifier(),
]
predictor_params = [
# None,
# None,
#{'n_iter':50, 'alpha': numpy.logspace(-3, 2) },
# {'name':'RBFSVC', 'n_iter':50,'C': numpy.logspace(-2, 6), 'gamma': numpy.logspace(-9, 3)},
{
#criacao o dicionario
#dict = []
#cont = 0
#for t in dataset.letra[:]:
# for w in t.split():
# dict.append(w)
#dict = set(dict) #dict agora contem todas as palavras existentes no dataset
#dict_array = list(dict) #precisa transformar pois o fit do label encoder nao aceita set
#separar o dataset em uma parte para fit e outra para predict
letras_train, letras_test, label_train, label_test = train_test_split(
dataset.letra, dataset.label, test_size=0.30, random_state=42)
#uso do LinearSVC
clf = SVC(C=1, loss='squared_hinge', penalty='l1', dual=False)
#print(dict_array)
#rotinas para alimentar o LabelEnconder
label_encoder = LabelEncoder()
int_encoded_fit = label_encoder.fit_transform(letras_train)
int_encoded_pred = label_encoder.fit_transform(letras_test)
#dict_encode = label_encoder.fit(dict_array)
#test_encode = []
#cont = 0
#while cont < 10:
# test_encode.append(dataset.letra[cont].split())
# cont += 1
from sklearn.datasets import make_blobs
import mglearn
import numpy as np
import matplotlib.pyplot as plt
from sklearn.svm.classes import LinearSVC
from Lib.idlelib.colorizer import color_config
X,y = make_blobs(random_state=42)
mglearn.discrete_scatter(X[:,0],X[:,1],y)
plt.xlabel("Feature 0")
plt.ylabel("Feature 1")
plt.legend(["Class 0","Class 1","Class 2"])
plt.show()
##在这个数据集上训练一个LinearSVC分类器
linear_svm = LinearSVC().fit(X,y)
print("Coefficient shape: ",linear_svm.coef_.shape)
print("Intercept shape: ",linear_svm.intercept_.shape)
#Coefficient shape: (3, 2)
#Intercept shape: (3,)
#coef_每行包含三个类别之一的系数向量，每列包含某个特征对应的系数值，包含两个特征
#intercept_是一维数组，保存每个类别的截距
#将这3个分类器给出的直线可视化
mglearn.discrete_scatter(X[:,0],X[:,1],y)
line = np.linspace(-15,15)
for coef,intercept,color in zip(linear_svm.coef_,linear_svm.intercept_,['b','r','g']):
plt.plot(line,-(line*coef[0]+intercept)/coef[1],c=color)
plt.ylim(-10,15)
plt.xlim((-10,8))
def setUp(self):
super(LinearSVCTest, self).setUp()
self.porter = Porter(language='ruby')
self._port_model(LinearSVC(C=1., random_state=0))
op.add_option('--use_available_classifiers', action='store_true', dest='available_classifiers',
help='Uses previously generated classifiers. If does not exists, new classifier models are \
generated.')
op.add_option("--random_state", action='store', type='int', dest='random_state',
help='Use random value of type int to reproduce the results.')
(opts, args) = op.parse_args()
if len(args) > 0:
op.error("this script takes no arguments.")
sys.exit(1)
if __doc__:
print(__doc__)
op.print_help()
clfs = (
(LinearSVC(), "Linear SVC"),
)
# total results from classifier
def calculateScore(result, n_folds):
print 'x' * 70
n_classifier = len(clfs)
for x in xrange(n_classifier):
a=0.0
p=0.0
r=0.0
f1=0.0
for i in xrange(x * n_folds, (x * n_folds) + n_folds):
a += result[i][0]
p += result[i][1]
r += result[i][2]
QENC_QUAL=False
QENC_DIFF=False
qenc_width = 33
n_classes = 3
FEAT_F33 = "F33"
n_users = 1000
max_runs = None #10000
percTest = 0.20
predictors = [
# DummyClassifier(strategy="stratified"),
# DummyClassifier(strategy="uniform"),
# BernoulliNB(),
LinearSVC(max_iter=100, class_weight="balanced"),
MLPClassifier(max_iter=100, nesterovs_momentum=True, early_stopping=True), #, activation="logistic"),
LogisticRegression(class_weight='balanced'),
# GaussianNB(),
]
predictor_params = [
# None,
# None,
# {'n_iter':50, 'alpha': numpy.logspace(-3, 2) },
{'n_iter':50,'C': numpy.logspace(-3, 2)},
{'n_iter':125,'hidden_layer_sizes':[(100,), (66,10)], 'learning_rate_init':[0.001, 0.01, 0.1], 'alpha': numpy.logspace(-6,2) },
{'n_iter':50,'C': numpy.logspace(-3, 2)},
# None,
]
return docs, t_docs, t_docsCategories
data = readData('hackerrank/documentClassification.txt')
X_train = np.array(data[1])
y_train = np.array(data[2])
X_test = np.array(data[0])
print("Extracting features from the training dataset using a sparse vectorizer")
#vectorizer = HashingVectorizer(stop_words='english', non_negative=True)
vectorizer = TfidfVectorizer(min_df=2,
ngram_range=(1, 2),
stop_words='english',
strip_accents='unicode',
norm='l2')
X_train = vectorizer.fit_transform(X_train)
#vectorizer = TfidfVectorizer(sublinear_tf=True, max_df=0.5,
# stop_words='english')
#X2_train = vectorizer.fit_transform(data_train.data)
X_test = vectorizer.transform(X_test)
nb_classifier = MultinomialNB().fit(X_train, y_train)
svm_classifier = LinearSVC().fit(X_train, y_train)
maxent_classifier = LogisticRegression().fit(X_train, y_train)
y_nb_predicted = nb_classifier.predict(X_test)
print(y_nb_predicted)
y_nb_predicted = svm_classifier.predict(X_test)
print(y_nb_predicted)
y_nb_predicted = maxent_classifier.predict(X_test)
print(y_nb_predicted)
def setUp(self):
super(LinearSVCTest, self).setUp()
self.porter = Porter(language='c')
self.set_classifier(LinearSVC(C=1., random_state=0))
train_r1 = time()
# prediction or testing
test_r0 = time()
predict = clf_rbf.predict(features_test)
test_r1 = time()
print "accuracy: ", clf_rbf.score(features_test, labels_test)
print "#################################"
print "tain time: ", round(train_r1 - train_r0, 3), "s"
print "prediction time: ", round(test_r1 - test_r0, 3), "s"
print "#################################"
'''
#SVC lib_linear
print("lib_linear")
clf_lib=LinearSVC()
# training
train_l0 = time()
clf_lib.fit(features_train, labels_train)
train_l1 = time()
# prediction or testing
test_l0 = time()
predict = clf_lib.predict(features_test)
test_l1 = time()
print "accuracy: ", clf_lib.score(features_test, labels_test)
print "#################################"
print "tain time: ", round(train_l1 - train_l0, 3), "s"
print "prediction time: ", round(test_l1 - test_l0, 3), "s"
def setUp(self):
super(LinearSVCPHPTest, self).setUp()
self.mdl = LinearSVC(C=1., random_state=0)
train_file = '/home/jvujjini/Kaggle/ForestCoverTypePrediction/train.csv'
test_file = '/home/jvujjini/Kaggle/ForestCoverTypePrediction/test.csv'
train_data = np.loadtxt(train_file, np.float32, delimiter=',')
test_data = np.loadtxt(test_file, np.float32, delimiter=',')
#training_data, training_label, test_data, test_label = train_data[:15000,:-1], data[:15000,-1], data[15000:,:-1], data[15000:,-1]
train_X = train_data[:, :-1]
train_y = train_data[:, -1]
test_X = test_data
print "starting..."
predict_label = OneVsRestClassifier(LinearSVC(random_state=0)).fit(
train_X, train_y).predict(test_X)
'''print "Started Training..."
clf.fit(train_X, train_y)
print "Done Training"
print "Started Predicting..."
predict_label = clf.predict(test_X)'''
output_file = '/home/jvujjini/Kaggle/ForestCoverTypePrediction/output.csv'
with open(output_file, 'w') as thefile:
print "File Opened..."
for item in predict_label:
thefile.write("%s\n" % item)
print "Success!"
def runModel(X, y, model_name):
nFolders = 5
accs = []
precs = []
recalls = []
F1s = []
n = X.shape[0]
for exp in range(0, nFolders):
print '\n\n============================================================================================\nexperiment', exp
### 2.1 split training and testing data
start = (int)((1 - (exp + 1) * 1.0 / nFolders) * n)
end = (int)((1 - exp * 1.0 / nFolders) * n)
#print n, start, end
X_train, y_train, X_test, y_test = splitTrainTest(X, y, start, end)
print 'Running', model_name
if model_name == 'SVM':
### 2.2 build classifier
clf = LinearSVC(penalty="l1", dual=False, tol=1e-7)
clf.fit(X_train, y_train)
### 2.3 predict
y_pred = clf.predict(X_test)
if model_name == 'SVM_new':
### 2.2 build classifier
clf = svm.SVC(C=1.0, gamma=1.0, class_weight='auto')
clf.fit(X_train, y_train)
### 2.3 predict
y_pred = clf.predict(X_test)
elif model_name == 'NaiveBayes':
clf = GaussianNB()
clf.fit(X_train.todense(), y_train)
y_pred = clf.predict(X_test.todense())
elif model_name == 'LogisticRegression':
clf = LogisticRegression(C=1.0, penalty='l1', tol=0.01)
clf.fit(X_train.toarray(), y_train)
y_pred = clf.predict(X_test.toarray())
else:
raise Exception("The model name is incorrect!!!")
### 2.4 eval
acc, prec, recall, F1 = eval(y_test, y_pred)
print 'Acc = ', acc
print 'Precision =', prec
print 'Recall=', recall
print 'F1 =', F1
accs.append(acc)
precs.append(prec)
recalls.append(recall)
F1s.append(F1)
print '\n\n\n'
print 'avg Acc = ', sum(accs) / len(accs)
print 'avg Precision = ', sum(precs) / len(precs)
print 'avg Recall = ', sum(recalls) / len(recalls)
print 'avg F1 = ', sum(F1s) / len(F1s)
return sum(accs) / len(accs), sum(precs) / len(precs), sum(recalls) / len(
recalls), sum(F1s) / len(F1s)
train_acc = metrics.accuracy_score(y_train, model.predict(x_train))
test_acc = metrics.accuracy_score(y_test, y_hat)
print u'训练集准确率：%.2f%%' % (100 * train_acc)
print u'测试集准确率：%.2f%%' % (100 * test_acc)
return t_train, t_test, 1 - train_acc, 1 - test_acc, name
#开始传提参数
clfs = [[RidgeClassifier(), 'Ridge'], [KNeighborsClassifier(), 'KNN'],
[MultinomialNB(), 'MultinomialNB'], [BernoulliNB(), 'BernoulliNB'],
[RandomForestClassifier(n_estimators=200), 'RandomForest'],
[SVC(), 'SVM'],
[
LinearSVC(loss='squared_hinge', penalty='l1', dual=False,
tol=1e-4), 'LinearSVC-l1'
],
[
LinearSVC(loss='squared_hinge', penalty='l2', dual=False,
tol=1e-4), 'LinearSVC-l2'
]]
#开始训练
result = []
for clf, name in clfs:
a = benchmark(clf, name)
result.append(a)
print '\n'
result = np.array(result)
genres = list(data_df.drop(['title', 'plot'], axis=1).columns.values)
data_x = data_df[['plot']].as_matrix()
data_y = data_df.drop(['title', 'plot'], axis=1).as_matrix()
stratified_split = StratifiedShuffleSplit(n_splits=2, test_size=0.33)
x_train, x_test, y_train, y_test = train_test_split(data_x,
data_y,
test_size=0.33,
random_state=42)
# transform matrix of plots into lists to pass to a TfidfVectorizer
train_x = [x[0].strip() for x in x_train.tolist()]
test_x = [x[0].strip() for x in x_test.tolist()]
stop_words = set(stopwords.words('english'))
## http://michelleful.github.io/code-blog/2015/06/20/pipelines/
## learn feature union to add more features (time, region)
pipeline = Pipeline([
('tfidf', TfidfVectorizer(stop_words=stop_words)),
('clf', OneVsRestClassifier(LinearSVC(), n_jobs=1)),
])
parameters = {
'tfidf__max_df': (0.25, 0.5, 0.75),
'tfidf__ngram_range': [(1, 1), (1, 2), (1, 3)],
"clf__estimator__C": [0.01, 0.1, 1],
"clf__estimator__class_weight": ['balanced', None],
}
grid_search(train_x, y_train, test_x, y_test, genres, parameters, pipeline)
train_r1 = time()
# prediction or testing
test_r0 = time()
predict = clf_rbf.predict(features_test)
test_r1 = time()
print "accuracy: ", clf_rbf.score(features_test, labels_test)
print "#################################"
print "tain time: ", round(train_r1 - train_r0, 3), "s"
print "prediction time: ", round(test_r1 - test_r0, 3), "s"
print "#################################"
'''
#SVC lib_linear
print("lib_linear")
clf_lib = LinearSVC()
# training
train_l0 = time()
clf_lib.fit(features_train, labels_train)
train_l1 = time()
# prediction or testing
test_l0 = time()
predict = clf_lib.predict(features_test)
test_l1 = time()
print "accuracy: ", clf_lib.score(features_test, labels_test)
print "#################################"
print "tain time: ", round(train_l1 - train_l0, 3), "s"
print "prediction time: ", round(test_l1 - test_l0, 3), "s"