def result():
if request.method == 'POST':
str1 = request.form['text1']
categories = [
'alt.atheism', 'soc.religion.christian', 'comp.graphics', 'sci.med'
]
twenty_train = fetch_20newsgroups(subset='train',
categories=categories,
shuffle=True,
random_state=42)
#filename = 'finalized_model.sav'
#clf = pickle.load(open(filename, 'rb'))
docs_new = list()
docs_new.append(str1)
print(docs_new)
proc_article = Pipeline(
['vect', CountVectorizer(), 'tfidf',
TfidfTransformer()])
proc_article._transform(docs_new)
predicted = clf.predict(proc_article)
for doc, category in zip(docs_new, predicted):
result = twenty_train.target_names[category]
return render_template("result.html", type=result)
def pipeline_transform(pipeline, X):
identity_pipeline = Pipeline(pipeline.steps[: -1] + [("estimator", None)])
return identity_pipeline._transform(X)
print("*** Transformed data : with LLE ***\n")
print(D)
#reduction de dimension avec la selection des varriables
#cette prepmier nous le reduit on un espace de 9 dimensions
Prepro.selectFeatures(X, y)
D.data['X_train'] = Prepro.transform(X)
D.data['X_valid'] = Prepro.transform(x_valid)
D.data['X_test'] = Prepro.transform(x_test)
estimators = [('imputer', Imputer()), ('scaler', MinMaxScaler())]
#puis passe au etapes suivantes qu on regroupe dans des piplines
pipe = Pipeline(estimators)
D.data['X_train'] = pipe.fit_transform(D.data['X_train'],
D.data['Y_train'])
D.data['X_valid'] = pipe._transform(D.data['X_valid'])
D.data['X_test'] = pipe._transform(D.data['X_test'])
print(
"\n*** Transformed data : selction des features avec LinearSVC des svm ***\n"
)
print(D)
#cette deuxieme methodes de selction nous permet de reduire la dimension de nos varriables à 10
Prepro.selectFeatures2(X, y)
D.data['X_train'] = Prepro.transform(X)
D.data['X_valid'] = Prepro.transform(x_valid)
D.data['X_test'] = Prepro.transform(x_test)
#on utilise le meme pipline que celui du test qui precede pour les autres traitements
pipe = Pipeline(estimators)
D.data['X_train'] = pipe.fit_transform(D.data['X_train'],
