Code

import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
from sklearn.naive_bayes import MultinomialNB
from sklearn.feature_extraction.text import CountVectorizer

#import dataset

spam_df= pd.read_csv("spam.csv")

#inspect data

spam_df

#group by

spam_df.groupby("Category").describe()

spam_df['spam']= spam_df['Category'].apply(lambda x:1 if x == 'spam' else 0)

# turn spam/ham into numerical data , creating a new column called "spam"
# lambda is a dummy function which stores a argument x from column Category and apply if else condition from "spam" or "not spam"

spam_df

# here spam column makes it easier to classify our data as ham or spam into numerical basis 0 or 1

'''---------------------------------------------------------------------------------------------------'''

#create a train test split

x_train,x_test,y_train,y_test = train_test_split(spam_df.Message, spam_df.spam)

# above x variable is going to be my feature as message
# by default test_size = 0.25 we can adjust it like 0.50 or 0.75

x_train

# features are word count
# labels are spam or not spam

x_train.describe()

# we've 4179 total emails in split 

# find word count and store data as numerical matrix 

cv = CountVectorizer()
x_train_count =cv.fit_transform(x_train.values)

# fit_transform used for transform all word counts into matrix

x_train_count          # numerical data

# sparse matrix has 4179 rows across this msgs 7529 are unique words 

x_train_count.toarray()

# train model

model = MultinomialNB()
model.fit(x_train_count,y_train)

# pre-test ham

email_ham = ["could you help me ?"]
email_ham_count = cv.transform(email_ham)
model.predict(email_ham_count)

# pre-test spam
email_spam = ["free"]
email_spam_count = cv.transform(email_spam)
model.predict(email_spam_count)

# test model
x_test_count = cv.transform(x_test)
model.score(x_test_count,y_test)