Detecting Spam Emails Using Tensorflow in Python

Spam messages are unsolicited or unwanted emails/messages sent in bulk to users. Detecting spam emails automatically helps prevent unnecessary clutter in users' inboxes.

In this article, we will build a spam email detection model that classifies emails as Spam or Ham (Not Spam) using TensorFlow, one of the most popular deep learning libraries.

Step 1: Import Required Libraries

Before we begin let’s import the necessary libraries: pandas, numpy, tensorflow, matplotlib, wordcloud, nltk for data processing, model building, and visualization.

Step 2: Load the Dataset

We’ll use a dataset containing labeled emails (Spam or Ham). Let’s load the dataset and inspect its structure. You can download the dataset from here:

Output:

This will give us a glimpse into the first few rows of the dataset. You can also check the shape of the dataset:

Output:

(5171,4)

The data contains 5171 rows and four columns.

Now, let's visualize the label distribution to get understanding of the class distribution:

Output:

Step 3: Balance the Dataset

We can clearly see that number of samples of Ham is much more than that of Spam which implies that the dataset we are using is imbalanced. To address the imbalance we’ll downsample the majority class (Ham) to match the minority class (Spam).

Output:

Step 4: Clean the Text

Textual data often requires preprocessing before feeding it into a machine learning model. Common steps include removing stopwords, punctuations, and performing stemming/lemmatization.

We’ll perform the following steps:

• Stopwords Removal
• Punctuations Removal
• Stemming or Lemmatization

Although removing data means loss of information we need to do this to make the data perfect to feed into a machine learning model.

Output:

Output:

The below function is a helper function that will help us to remove the stop words.

Output:

Visualization Word Cloud

A word cloud is a text visualization tool that help's us to get insights into the most frequent words present in the corpus of the data.

Output:

Step 6: Tokenization and Padding

Machine learning models work with numbers, so we need to convert the text data into numerical vectors using Tokenization and Padding.

1. Tokenization: Converts each word into a unique integer.
2. Padding: Ensures that all text sequences have the same length, making them compatible with the model.

Step 7: Define the Model

We will build a deep learning model using a Sequential architecture. This model will include:

• Embedding Layer: Learns vector representations of words.
• LSTM Layer: Captures patterns in sequences.
• Fully Connected Layer: Extracts relevant features.
• Output Layer: Predicts whether an email is spam or not.

Output:

Model: "sequential"
__________________________________________________
Layer (type) Output Shape Param #
========================================================
embedding (Embedding) (None, 100, 32) 1274912

lstm (LSTM) (None, 16) 3136

dense (Dense) (None, 32) 544

dense_1 (Dense) (None, 1) 33

========================================================
Total params: 1,278,625
Trainable params: 1,278,625
Non-trainable params: 0
__________________________________________________

Step 8: Train the Model

We train the model using EarlyStopping and ReduceLROnPlateau callbacks. These callbacks help stop the training early if the model’s performance doesn’t improve and reduce the learning rate to fine-tune the model.

Output:

After training, we evaluate the model on the test data to measure its performance.

Output: 

Test Loss: 0.1202
Test Accuracy: 0.9700

Thus, the training accuracy turns out to be 97% which is quite satisfactory.

Having trained our model, we can plot a graph depicting the variance of training and validation accuracies with the no. of epochs.

Output:

By following these steps, we have successfully built a machine learning model that can classify emails as spam or ham. With further optimization, this model can be fine-tuned to improve its performance even more.