def Xgboost():
train_data, train_label = genKeyWords("../data/cnews.train.txt")
vectors = CountVectorizer(max_df=0.2)
tfidf = TfidfTransformer(use_idf=True)
bst=XGBClassifier(n_jobs=10,max_depth=55,objective='multi:softmax',num_class=10,subsample=0.4,reg_lambda=0.8)
pipline=Pipeline([("vectors",vectors),("tfidf",tfidf),("bst",bst)])
pipline.fit(train_data,train_label)
joblib.dump(pipline,"./model/XGB.m")
test_data, test_label = genKeyWords("../data/cnews.test.txt")
predicted = pipline.predict(test_data)
print('Xgboost', np.mean(predicted == test_label))
def svc():
count=CountVectorizer(max_df=0.2,max_features=None)
tfidf=TfidfTransformer(use_idf=False)
_svc=SVC(C=0.99,kernel='linear')
train_data, train_label=genKeyWords("../data/cnews.train.txt")
test_data,test_label=genKeyWords("../data/cnews.test.txt")
pipline=Pipeline([("count",count),("tfidf",tfidf),("svc",_svc)])
pipline=pipline.fit(train_data,train_label)
joblib.dump(pipline, "./SVM.m")
predicted = pipline.predict(test_data)
print('SVC', np.mean(predicted == test_label))
def Bayes(mode='mul'):
if mode=='mul':
model=MultinomialNB()
elif mode=='gau':
model=GaussianNB()
elif mode=='bern':
model=BernoulliNB()
else:
raise ValueError('没有该模式，请填写以下mode，\n mul==>MultinomialNB \ngau==>GaussianNB \n bern==>BernoulliNB')
train_data, train_label = genKeyWords("../data/cnews.train.txt")
vectors = CountVectorizer()
tfidf=TfidfTransformer()
pipline=Pipeline([("vectors",vectors),("tfidf",tfidf),("bayes",model)])
pipline.fit(train_data,train_label)
test_data, test_label = genKeyWords("../data/cnews.test.txt")
predicted = pipline.predict(test_data)
joblib.dump(pipline, "./%s_bayes.m"%mode)
print('naive_bayes', np.mean(predicted == test_label))
def Knn():
train_data, train_label=genKeyWords("../data/cnews.train.txt")
vectors=CountVectorizer()
tfidf=TfidfTransformer()
clf=KNeighborsClassifier(n_neighbors=5,n_jobs=-1)
pipline=Pipeline([("vectors",vectors),("tfidf",tfidf),("clf",clf)])
# parameters={'clf__n_neighbors':list(range(5,20,2))}
# grid_search = GridSearchCV(pipline, parameters, n_jobs=-1, verbose=1)
# grid_search.fit(train_data,train_label)
# best_parameters = grid_search.best_estimator_.get_params()
# for param_name in sorted(parameters.keys()):
# print("\t%s: %r" % (param_name, best_parameters[param_name]))
pipline.fit(train_data,train_label)
test_data,test_label=genKeyWords("../data/cnews.test.txt")
predicted = pipline.predict(test_data)
joblib.dump(pipline,"./knn.m")
print('KNeighborsClassifier', np.mean(predicted == test_label))
def DTrees():
train_data, train_label = genKeyWords("../data/cnews.train.txt")
vectors = CountVectorizer(max_df=0.6)
tfidf = TfidfTransformer(use_idf=False)
tree=DecisionTreeClassifier(criterion="entropy",max_depth=20)
pipline=Pipeline([("vectors",vectors),("tfidf",tfidf),("tree",tree)])
params={"tree__max_depth":list(range(75,105,5))}
accuracy=make_scorer(accuracy_score)
gridsearch=GridSearchCV(pipline,params,n_jobs=10,scoring=accuracy)
gridsearch.fit(train_data,train_label)
best_parameters = gridsearch.best_estimator_.get_params()
for param_name in sorted(params.keys()):
print("\t%s: %r" % (param_name, best_parameters[param_name]))
def K_Means(minibatch):
train_data, train_label = genKeyWords("../data/cnews.train.txt")
vectors = CountVectorizer()
tfidf = TfidfTransformer()
if minibatch:
km = MiniBatchKMeans(n_clusters=10, init='k-means++', n_init=1,
init_size=1000, batch_size=1000, verbose=False)
else:
km = KMeans(n_clusters=10, init='k-means++', max_iter=300, n_init=1,
verbose=False)
pipline=Pipeline([("vectors",vectors),("tfidf",tfidf),("kmeans",km)])
pipline.fit(train_data)