How to do predict a new sms to be spam or not?

Question

I have trained a model for spam classification -

This is my code -

X_train, X_test, y_train, y_test = train_test_split(data['text'], data['label'], test_size = 0.4, random_state = 1)

cv = CountVectorizer()
cv.fit(X_train)
cv.fit(X_test)
X_train = cv.transform(X_train)
X_test = cv.transform(X_test)

model = LogisticRegression(solver='lbfgs')
model.fit(X_train, y_train)

After that I had also completed testing, it gives me an accuracy of about 97% .

Now I want to add predict a new SMS/Email to be spam or not. What I am doing is -

new = 'Hey there you got a sale here on website'    
new = cleanText(new)
cv.fit([new])
new = cv.transform([new])
model.predict(new)

It gives me an error

ValueError: X has 4 features per sample; expecting 4331

Please tell me where I am going wrong?

Answer 1

You are refiting the CountVectorizer during prediction. Just remove the line

cv.fit([new])

and it should work.

Answer 2

You must never fit the test set, you have to use it only to predict so you can evaluate your model, same with new predictions, only fit training set.