def tune_parameters(self, file):
data = dr(file).read_data(precision = self.precision)
features = data[:, 2:-1]
labels = data[:, -1].astype(int)
fT, ft, lT, lt = train_test_split(features, labels, test_size = 0.5, random_state = 0)
parameters = [{'C': [1, 5, 10, 50, 100, 500, 1000], 'tol': [1e-2, 1e-3, 1e-4, 1e-5, 1e-6],
'class_weight': ['balanced', None]}]
clf = GridSearchCV(LogisticRegression(), parameters, cv = 2, scoring = make_scorer(precision_score, pos_label = 1))
clf.fit(fT, lT)
print("-----\nBest parameters set found for Logistic Regression tuning precision:\n-----")
print(clf.best_params_)
print(classification_report(lt, clf.predict(ft)))
def validate_proba(self, file, proba_tol):
data = dr(file).read_data(precision=self.precision)
features = data[:, 2:-1]
labels = data[:, -1].astype(int)
result = []
k_fold = KFold(n_splits=2)
for train, test in k_fold.split(features):
self.train(features[train], labels[train])
preds = self.proba_util(features[test], proba_tol)
result.append(precision_score(labels[test], preds, pos_label=1))
dr.print_report(self.debug, preds,
labels[test]) #Print Classification Report
return result
def predict(self, file):
data = dr(file).read_data(precision=self.precision)
features = data[:, 0:-1]
labels = data[:, -1].astype(int)
print('[INFO] ', np.count_nonzero(labels != 0),
'links present in training data, out of ', len(labels))
k_fold = KFold(n_splits=2)
for train, test in k_fold.split(features):
self.train(features[train], labels[train])
preds = self.ada.predict(features[test])
print('[INFO] ', np.count_nonzero(preds != 0),
' links retrieved by model')
print('[INFO] Succesfully predicted ',
np.count_nonzero(preds == labels[test]), ' links out of ',
len(labels[test]))
from utils.data_reader import DataReader as dr
# Build a forest and compute the feature importances
forest = RandomForestClassifier(n_estimators=17,
max_depth=5,
criterion='entropy',
random_state=0,
class_weight='balanced')
forest = AdaBoostClassifier(n_estimators=4000,
learning_rate=1,
algorithm='SAMME.R')
forest = AdaBoostClassifier(n_estimators=50,
learning_rate=1,
algorithm='SAMME.R')
data = dr('corpus_scores\\v2_5_raw_inv.txt').read_data(precision=-1)
features = data[:, 2:-1]
labels = data[:, -1].astype(int)
forest.fit(features, labels)
importances = forest.feature_importances_
std = np.std([tree.feature_importances_ for tree in forest.estimators_],
axis=0)
indices = np.argsort(importances)[::-1]
# Print the feature ranking
print("Feature ranking:")
for f in range(features.shape[1]):
print("%d. feature %d (%f)" % (f + 1, indices[f], importances[indices[f]]))
