How to classify a new email as spam/not spam?

Question

I am working on a small exercise for determining if an email is spam or not. My dataset is the following:

                       Email                                                   Spam
0   Hi, I am Andrew and I want too buy VIAGRA                                   1
1   Dear subscriber, your account will be closed                                1
2   Please click below to verify and access email restore                       1
3   Hi Anne, I miss you so much! Can’t wait to see you                          0
4   Dear Professor Johnson, I was unable to attend class today                  0
5   I am pleased to inform you that you have won our grand prize.               1
6   I can’t help you with that cuz it’s too hard.                               0
7   I’m sorry to tell you but im sick and will not be able to come to class.    0
8   Can I see an example before all are shipped or will that cost extra?        0
9   I appreciate your assistance and look forward to hearing back from you.     0

where 1 means spam, 0 not spam. What I have tried is the following:

#Tokenization 

def fun(t):

# Removing Punctuations
remove_punc = [c for c in text if c not in string.punctuation]
remove_punc = ''.join(remove_punc)

# Removing StopWords
cleaned = [w for w in remove_punc.split() if w.lower() not in stopwords.words('english')]

return cleaned

So I applied the function: df['Email'].apply(fun). Then I converted the text into a matrix as follows:

from sklearn.feature_extraction.text import CountVectorizer
mex = CountVectorizer(analyzer= fun).fit_transform(df['Email'])

and split the dataset into train and test:

X_train, X_test, y_train, y_test = train_test_split(mex, df['Email'], test_size = 0.25, random_state = 0)

I applied a classifier (I would apply the logistic regression to determine if an email is spam or not, but I have currently only used Naive Bayes:

from sklearn.naive_bayes import MultinomialNB

classifier = MultinomialNB()
classifier.fit(X_train, y_train)

Finally I applied the classifier first to the train set, then to the test set:

from sklearn.metrics import classification_report,confusion_matrix, accuracy_score
from sklearn.metrics import classification_report,confusion_matrix, accuracy_score

pred = classifier.predict(X_test)

print(classification_report(y_test ,pred ))
print('Confusion Matrix: \n', confusion_matrix(y_test,pred))
print()
print('Accuracy: ', accuracy_score(y_test,pred))

The code works but I would like to know how to visually see with an example of new email if this has label 1 or 0. For example: if I have a new email 'Hi, my name is Christopher and I like VIAGRA', how could I determine the label/class?

I feel I am missing something or probably I am following a wrong way to demonstrate this.

My question is the following:

Given this new email: Hi, my name is Christopher and I like VIAGRA, how can I see if this is spam or not? I have thought about classification but probably my approach is wrong. I would like to have something like:

Email                                        Spam 
... 
Hi, my name is Christopher and I like VIAGRA 1

as this is very similar to the email 'Hi, I am Andrew and I want too buy VIAGRA' (if included in the train set or correctly predicted in the test set).

Maybe what I want to do requires only tf-idf algorithm or a different approach. Any advice will be appreciated.

Answer 1

I modified your code so the code runs as a block and is setup to predict new data:

import string

from nltk.corpus import stopwords
import pandas as pd
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report,confusion_matrix, accuracy_score
from sklearn.naive_bayes import MultinomialNB

#  Define training data
df = pd.DataFrame(data={'Email': [
"Hi, I am Andrew and I want too buy VIAGRA",
"Dear subscriber, your account will be closed",
"Please click below to verify and access email restore",
"Hi Anne, I miss you so much! Can’t wait to see you",
"Dear Professor Johnson, I was unable to attend class today",
"I am pleased to inform you that you have won our grand prize.",
"I can’t help you with that cuz it’s too hard.",
"I’m sorry to tell you but im sick and will not be able to come to class.",
"Can I see an example before all are shipped or will that cost extra?",
"I appreciate your assistance and look forward to hearing back from you.",], 
'Spam': [1, 1, 1, 0, 0, 1, 0, 0, 0, 0]})

def fun(text):    
    # Removing Punctuations
    remove_punc = [c for c in text if c not in string.punctuation]
    remove_punc = ''.join(remove_punc)

    # Removing StopWords
    cleaned = [w for w in remove_punc.split() if w.lower() not in stopwords.words('english')]

    return cleaned

# Create a vectorizer object to enable both fit_transform and just transform
vectorizer = CountVectorizer(analyzer=fun)
X = vectorizer.fit_transform(df['Email'])

X_train, X_test, y_train, y_test = train_test_split(X, df['Spam'], test_size = 0.25, random_state = 0)

classifier = MultinomialNB()
classifier.fit(X_train, y_train)

pred = classifier.predict(X_test)

print(classification_report(y_test ,pred ))
print('Confusion Matrix: \n', confusion_matrix(y_test,pred))
print()
print('Accuracy: ', accuracy_score(y_test,pred))

Here is how to predict on new data:

# Given a new email
new_email = "Hi, my name is Christopher and I like VIAGRA"

# Apply the same preprocessing steps and transformation
X_new = vectorizer.transform([fun(new_email)])

# Predict new email with already trained classifier
classifier.predict(X_new)