def LogReg(n=5):
X, y = feature_extraction.pre_process_comments() # get the data
X = SelectKBest(chi2, k=17000).fit_transform(X, y)
'''
LogReg_clf = Pipeline([
('feature_selection', SelectFromModel(LinearSVC(penalty="l1", dual=False, multi_class='crammer_singer', max_iter=5000), threshold="0.8*mean", max_features=50000)),
('classification', LogisticRegression(penalty='l1', solver='saga'))
])
'''
LogReg_clf = LogisticRegression(penalty='l1', dual=False)
print("before fitting")
LogReg_cv = cross_val_score(LogReg_clf, X, y, cv=n)
LogReg_clf.fit(X, y)
LogReg_pred = LogReg_clf.predict(X)
LogReg_acc = accuracy_score(y_pred=LogReg_pred, y_true=y)
print("cv acc of LogReg is ", LogReg_cv.mean())
print("acc of logreg is ", LogReg_acc)
return LogReg_clf
def Bernoulli(n=5):
X, y = feature_extraction.pre_process_comments() # get the data
X = SelectKBest(chi2, k=17000).fit_transform(X, y)
Bernoulli_clf = BernoulliNB()
Bernoulli_cv = cross_val_score(Bernoulli_clf, X, y, cv=n)
Bernoulli_clf.fit(X, y)
Bernoulli_pred = Bernoulli_clf.predict(X)
Bernoulli_acc = accuracy_score(y_true=y, y_pred=Bernoulli_pred)
print("acc of BNB(sklearn) is ", Bernoulli_acc)
print("cv acc of BNB(sklearn) is ", Bernoulli_cv.mean())
return Bernoulli_clf
def MultiNB_Kaggle():
X, y, test, test_ID = feature_extraction.pre_process_comments(
) # get the data
k_best = SelectKBest(chi2, k=17000)
X = k_best.fit_transform(X, y)
MultiNB_clf = MultinomialNB(alpha=0.22)
test_X = k_best.transform(test)
MultiNB_clf.fit(X, y)
MultiNB_pred = MultiNB_clf.predict(test_X)
np.savetxt('predict.csv',
np.array([test_ID, MultiNB_pred]).transpose(),
delimiter=',',
fmt='%s',
header='Id, Category')
return MultiNB_clf
def MultiNB(n=5):
X, y = feature_extraction.pre_process_comments() # get the data
X = SelectKBest(chi2, k=17000).fit_transform(X, y)
MultiNB_clf = MultinomialNB(alpha=0.22)
MultiNB_cv = cross_val_score(MultiNB_clf, X, y, cv=n)
MultiNB_clf.fit(X, y)
MultiNB_pred = MultiNB_clf.predict(X)
MultiNB_acc = accuracy_score(y_pred=MultiNB_pred, y_true=y)
print("cv acc of multiNB is ", MultiNB_cv.mean())
print("acc of multiNB is ", MultiNB_acc)
return MultiNB_clf
def LSVC(n=5):
X, y = feature_extraction.pre_process_comments() # get the data
X = SelectKBest(k=17000).fit_transform(X, y)
LSVC_clf = LinearSVC(penalty='l1', dual=False)
#LSVC_clf = BaggingClassifier(base_estimator=LSVC_clf, max_samples=0.6, max_features=12000, n_estimators=24, random_state=1)
print("before fitting")
LSVC_cv = cross_val_score(LSVC_clf, X, y, cv=n)
LSVC_clf.fit(X, y)
LSVC_pred = LSVC_clf.predict(X)
LSVC_acc = accuracy_score(y_pred=LSVC_pred, y_true=y)
print("cross validation accuracy", LSVC_cv.mean())
print("training accuracy", LSVC_acc)
return LSVC_clf
def ensemble(n=5):
X, y = feature_extraction.pre_process_comments() # get the data
X = SelectKBest(k=17000).fit_transform(X, y)
LogReg_clf = LogisticRegression(penalty='l1', dual=False)
MNB_clf = MultinomialNB(alpha=0.22)
LinearSVC_clf = LinearSVC(penalty='l1', dual=False)
Ensemble_clf = VotingClassifier(estimators=[('lg', LogReg_clf),
('mnb', MNB_clf)],
voting='soft')
Ensemble_cv = cross_val_score(Ensemble_clf, X, y, cv=n)
Ensemble_clf.fit(X, y)
score = Ensemble_clf.score(X, y)
print("cv acc of ensemble is ", Ensemble_cv.mean())
print("acc of ensemble is ", score)
return Ensemble_clf
from sklearn.model_selection import cross_val_score
from sklearn.tree import DecisionTreeClassifier
from sklearn.svm import LinearSVC
from sklearn.pipeline import Pipeline
import tensorflow as tf
from tensorflow.keras import layers
import feature_extraction
from sklearn.naive_bayes import MultinomialNB
from sklearn.ensemble import RandomForestClassifier, VotingClassifier
import models.models as models
from sklearn.feature_selection import SelectKBest, chi2, VarianceThreshold
from sklearn.feature_selection import RFE
from sklearn.svm import SVR
from sklearn.preprocessing import LabelEncoder
X, y = feature_extraction.pre_process_comments() # get the data
X = SelectKBest(chi2, k=17000).fit_transform(X, y)
LogReg_clf = LogisticRegression(penalty='l1', dual=False)
MNB_clf = MultinomialNB(alpha=0.3)
LinearSVC_clf = LinearSVC(penalty='l1', dual=False)
Ensemble_clf = VotingClassifier(estimators=[('lg', LogReg_clf),
('mnb', MNB_clf)],
voting='soft')
Ensemble_cv = cross_val_score(Ensemble_clf, X, y, cv=3)
Ensemble_clf.fit(X, y)