A Seminar Report

On

EMAIL SPAM DETECTION USING

MACHINE LEARNING

By

PATIL PRAJWAL PRABHAKAR

Roll No: 305B052

TE (Computer Engineering)

Under the guidance of

Prof N.G.Bhojne

Department of Computer Engineering

Sinhgad College of Engineering,
Vadgaon (Bk.), Pune-411041
Accredited by NAAC

Affiliated to Savitribai Phule Pune University, Pune

2024-2025
 Date:

CERTIFICATE
This is to certify that the Seminar Report

EMAIL SPAM DETECTION USING

MACHINE LEARNING
Submitted by

PATIL PRAJWAL PRABHAKAR

Has successfully completed her Seminar Report and Presentation under the supervision of respected
Prof N.G.Bhojne for the partial fulfillment of Third Year of Bachelor of Engineering, Computer
Engineering of Savitribai Phule Pune University.

Prof. N.G.Bhojne Dr. M. P. Wankhade

Seminar Guide Head, Computer Engineering Department

Dr. S. D. Lokhande,
Principal,
Sinhgad College of Engineering, Pune

Internal Examiner External Examiner

I
 ACKNOWLEDGEMENT

We would like to express our gratitude and appreciation to all those who gave us the possibility
to complete this report. Special thanks to our beloved guide Prof. N.G.Bhojne whose help, stimulating
suggestions and encouragements helped us in all time of fabrication process and in writing this report.
We also sincerely thank for the time spent for correcting our many of the mistakes.
We would like to acknowledge with much appreciation that the crucial role of the staff in Computer
Engineering Department, who gave us the permission to use the lab equipment and also the machines
and to design the report and the other stuffs.
We are equally grateful to Dr. M. P. Wankhade HOD of Computer Engineering Department for his
constant support and inspiration during this complete journey.
We would like to express our obligation to our respected Principal Sir, Dr. S. D. Lokhande for
providing excellent facilities for research work.
Our heart-fell thanks to all our beloved friends for their unforgettable help and moral support.

Name of Student
Ms. Patil Prajwal

II
CONTENTS

Certificate .............................................................................................................................................. I
Acknowledgement ............................................................................................................................... II
Contents ............................................................................................................................................. III
List of Figures .................................................................................................................................... III
Index .................................................................................................................................................. IV
Abstract ................................................................................................................................................V

LIST OF FIGURES

1.1 Introduction ........................................................................................................................2

1.2 Motivation ..........................................................................................................................4
2.1.1 Working of Neutral Network ...........................................................................................7
2.1.1 Features of Project ...........................................................................................................8
4.1 System Architecture ...........................................................................................................13
5.2.1Working............................................................................................................................19
5.2.2 Output Screen ................................................................................................................19

III
INDEX

1. INTRODUCTION ............................................................................................................1
1.1 Introduction to Proejct ...........................................................................................1
1.2 Problem Statement ................................................................................................3
1.3 Objective ...............................................................................................................3
1.4 Motivation ............................................................................................................4

2. BACKGROUND ..............................................................................................................5
2.1 Overview of System .......................................................................................6
2.1.1 Methodology ............................................................................................ 7-8

3. LITERATURE SURVEY ...............................................................................................9

3.1 Literature Survey ................................................................................................10
3.2 Summary .............................................................................................................11

4. SOFTWARE REQUIREMENT SPECIFICATION ...................................................12

4.1 Introduction to System Engineering ...................................................................13
4.2 System Requirements .........................................................................................13
4.2.1 Hardware Requirements ..................................................................................13
4.2.2 Software Requirements.....................................................................................13
4.3 Functional Requirements ....................................................................................14

5. RESULTS AND DISCUSSION .....................................................................................15

5.1 Results (SCREEN-SHOTS OF THE RESULT) .......................................... 17-18
5.1.1 Result Analysis .........................................................................................19

6. CONCLUSION & REFERENCES ...............................................................................20

IV
 ABSTRACT
This report provides an in-depth analysis of Email spam detection using machine learning and its
fundamental role in revolutionizing modern business operations. Email spam classification is a critical
task in today's digital world, where the amount of spam emails has increased dramatically. In this
project, we propose to use machine learning (ML) and natural language processing (NLP) techniques
to classify email messages as either spam or legitimate. The project aims to develop an efficient spam
classifier that can accurately identify and filter spam emails from legitimate ones.

The dataset used in this project will consist of a large number of email messages with their
corresponding labels (spam/ham). We will use NLP techniques such as tokenization, stop word
removal, stemming, and feature extraction to preprocess the text data and extract relevant features. We
will evaluate several ML algorithms such as Naive Bayes, Support Vector Machines (SVMs), and
Neutral network to determine the best model for spam classification. We will also perform hyper
parameter tuning to optimize the model's performance. The accuracy of the classifier will be measured
using evaluation metrics such as precision, recall, and F1-score.

The project's outcomes will include a spam classifier model that can be integrated into an email system
to automatically filter spam emails, improving email security and productivity. Additionally, the project
will contribute to the advancement of NLP and ML techniques for email spam classification.

Keywords: - Ham/spam, Natural Language Processing, Machine Learning, Online Platform, Email.

V
 CHAPTER 1
INTRODUCTION
1.1 INTRODUCTION TO PROJECT
Spam emails are a major problem, making up over half of all email traffic. They waste time,
storage space, and bandwidth. Spammers use various techniques to avoid detection, including
fake sender addresses and hidden information in images.

To fight spam, email providers use a combination of techniques. Machine learning is the most
effective approach, where algorithms learn to identify spam by analyzing massive datasets of
emails. This is more efficient than knowledge engineering, which requires manually creating
rules that need constant updates.

Common machine learning algorithms used for spam filtering include Naive Bayes, Support
Vector Machines, Neural Networks. This paper reviews the evolution of spam filtering
techniques, the architectures of spam filters, and different machine learning algorithms used by
Gmail, Yahoo Mail, and Outlook. It also highlights open research areas and suggests directions
for developing new techniques to combat future spam variants.

The paper concludes by highlighting the need for continued research in machine learning
techniques to address the evolving nature of spam. It recommends exploring algorithms like Deep
Learning and Genetic Algorithms, which have shown promise in other domains but have not been
fully explored for spam filtering.

In summary, this research paper presents a detailed exploration of our machine learning for email
spam detection using machine learning. offering insights into its potential to transform bussiness
practices and support a wide range of user needs.
1.2 PROBLEM STATEMENT

Email spam has become a significant problem in today's digital age, posing challenges for
individuals, businesses, and organizations alike. Spam emails are unsolicited messages that flood
inboxes, wasting valuable time and resources while potentially exposing users tomalicious
content or scams. To combat this issue, machine learning techniques have emerged as powerful
tools for email spam detection.
Spam detection using machine learning aims to develop efficient and accurate algorithms to
identify and filter unsolicited, unwanted emails, commonly known as spam. The primary
challenge lies in the ever-evolving nature of spam techniques, making it difficult to create static
rules for detection. Machine learning offers a dynamic approach by learning from vast datasets
of labeled emails, adapting to new spam patterns, and improving detection accuracy over time.
The objective of email spam detection is to accurately classify incoming emails as either
legitimate (ham) or spam. Traditional rule-based approaches have limited effectiveness due to
the constantly evolving nature of spam. Machine learning offers a more dynamic and adaptable
approach by leveraging patterns and features extracted from large email datasets. Machine
learning algorithms can learn from labelled email datasets to build models capable of recognizing
patterns indicative of spam. These models can then be used to automatically classify new, unseen
emails. By analyzing various email attributes such as sender information,subject line, content,
and embedded URLs, machine learning algorithms can identify spam characteristics and make
accurate predictions.
1.3 OBJECTIVE

o Accurate Spam Classification: Develop machine learning models that can accurately
distinguish between legitimate emails and spam, minimizing false positives and false
negatives.

o Real-time Detection: Create algorithms capable of detecting spam emails in real-time,

preventing them from reaching users' inboxes and causing inconvenience.

o Adaptability to Evolving Spam Tactics: Design models that can adapt to new spam techniques
and patterns, ensuring ongoing effectiveness against sophisticated spammers.
 o Efficiency and Scalability: Develop algorithms that are computationally efficient and can
handle large volumes of emails, making them suitable for real-world deployment.

o Privacy Preservation: Ensure that spam detection methods do not compromise user privacy
by avoiding the collection or storage of sensitive information.

1.4 MOTIVATION
The increasing prevalence of spam emails poses a significant challenge to individuals and
organizations alike. Spam not only wastes time and resources but also poses security risks, as it can be
used to spread malware, phishing attacks, and other malicious activities. To effectively combat spam,
traditional rule-based methods have limitations, as spammers constantly evolve their techniques.
Machine learning offers a promising solution to this problem. By leveraging the power of algorithms to
learn from vast datasets of labeled emails, machine learning models can adapt to new spam patterns and
improve their detection accuracy over time. This dynamic approach is essential to keep pace with the ever-
changing landscape of spam.
Additionally, machine learning can help address the scalability challenges associated with spam filtering.
Traditional methods often struggle to handle the massive volumes of emails generated daily. Machine
learning algorithms, on the other hand, can be designed to efficiently process large datasets, making them
suitable for real-world deployment.
CHAPTER 2
BACKGROUND
2.1 OVERVIEW OF SYSTEM
2.1.1 METHODOLOGY
There are two conventional types of neural networks that are usually implied whenever Artificial
neutral network (ANN) is used. They are the perceptron and the multilayer perceptron. This section
will attempt to explain the perceptron algorithm and its application to email spam filtering. Below
is a perceptron algorithm which is a standard Neural Network algorithm. The perceptron assists in
locating a linear function of the attribute vector f(x) > wT x + b such that f(x) > 0 for vectors of one
group , and f(x) < 0 for vectors of other group. Also, w ¼ (w1, w2,…wm) are the weights of the
function, and b is the supposed bias. The groups can be given the numbers þ1 and -1, so search for
a function d(x) ¼ sign (wT x þ b) is carried out. The perceptron learning begins by randomly
selecting parameters (w0,b0) of the resolution and repeatedly bringing them up-to-date. A training
sample (x,c) is selected at the nth iteration of the algorithm to the extent that the present decision
function now group it as incorrect (i.e. sign (wnx þ bn) 6¼ c). The rule depicted by Eq. (14) below
is used in updating the parameters (wn, bn):In the proposed methodology as shown in Fig. 1, front
end processes provide different controls/components like tools, operators and procedures to create
visual model or workflow. At the backend processes with respect to the components generate the
source code of the visual model. Validation of the component connectivity gives correct visual
model of the required system or application. The proposed system has functionality to execute the
applications without using optimization techniques as well as with using optimization techniques.
Executable application file is generated for the user at the end.

wnþ1 ¼ wn þ cx bnþ1 ¼ bn þ
The criteria for terminating the algorithm is that a decision function must be located which
accurately categorises all the training samples into different groups. The algorithm below is based
on this explanation that was just given [66]. There are times when the training data cannot be
separated linearly, in such cases the wisest action to take is to terminate the training algorithm
once the number of data that are erroneously classified is sufficiently small [67]. The algorithm
below represent the algorithm for a Perceptron Neural Network for email spam classification:FE =
{Tools C Connection, Tools C Property, Tools C Location}.
What is Neural networks?

Neural networks are computational models inspired by the structure and function of the human
brain. They are composed of interconnected nodes, often referred to as neurons, organized into
layers. These neurons process information by taking weighted sums of their inputs and applying an
activation function to determine their output.
Neural networks, though powerful for machine learning, are not commonly used for spam detection
due to the prevalence of Naive Bayes classifiers. However, recent studies suggest that incorporating
neural networks into spam filters can significantly improve their accuracy. Google's experience with
Gmail spam filters demonstrates this, where the accuracy increased from 99.5% to 99.9% after
adding neural networks. More research is needed to explore the potential of neural networks in spam
detection, focusing on network design, momentum, and learning rate to optimize their effectiveness
across various datasets.

Figure 2.1.1: Working of EMAIL SPAM detection

Key Features of Low Code Development Platform

A. Nonlinearity: Neural networks can effectively model complex nonlinear relationships between
features in email data, which are often present in spam messages. This allows them to capture subtle
patterns that may be difficult for linear models to detect.
B. Adaptability: Neural networks can learn from large datasets of labeled emails, adapting to new
spam tactics and evolving patterns over time. This adaptability is crucial for combating the ever-
changing nature of spam.
C. Feature Learning: Neural networks can automatically learn relevant features from email data,
eliminating the need for manual feature engineering. This can save time and effort, especially when
dealing with large and complex datasets.
D. Parallel Processing: Neural networks can be implemented on parallel hardware, such as GPUs,
to accelerate training and inference. This is particularly important for real-time spam detection,
where quick responses are essential.

Figure 2.1.2: Working of Neutral network

 CHAPTER 3
RELATED WORK
3.1 LITERATURE SURVEY
Rani, R., & Nasa, R. (2021). A comparative analysis of machine learning algorithms for spam email
detection. In Proceedings of the International Conference on Advances in Computing and
Communication Engineering (pp. 561-573). Springer, Singapore. This study performs a comparative
analysis of machine learning algorithms for spam email detection. It evaluates the performance of
algorithms such as Naive Bayes, decision trees, and K-nearest neighbours.

Ahirwal, A., & Kaushik, A. (2021). A comprehensive review of email spam detection techniques
using machine learning. In Proceedings of the International Conference on Advanced
Computational and Communication Paradigms (pp. 77-85). Springer, Singapore. This paper
presents a comprehensive review of email spam detection techniques using machine learning. It
covers various algorithms, feature extraction methods, and evaluation metrics used in the field.

Saini, R., & Kumar, R. (2020). Comparative study of machine learning techniques forspam email
detection. In Proceedings of the International Conference on Computational Intelligence and
Communication Technology (pp. 257-267). Springer, Singapore. This research compares various
machine learning techniques for spam email detection. It provides insights into the performance of
algorithms such as Naive Bayes, SVM, and random forests.

Salam, A., Al-Ayyoub, M., Aljawarneh, S., Jararweh, Y., & Gupta, B. (2020). Email spam detection
using machine learning: A comparative study. IEEE Access, 8, 78782-78796. This study conducts a
comparative analysis of different machine learning algorithms for email spam detection. It evaluates
the performance of algorithms such as Naive Bayes, SVM, and decision trees.

Mishra, A., Joshi, R. C., & Gaur, M. S. (2020). A comprehensive review on email spam detection
techniques using machine learning. In Proceedings of the International Conference on Advances in
Computing and Data Sciences (pp. 129-140). Springer, Singapore. This research presents a
comprehensive review of email spam detection techniques using machine learning. It covers various
algorithms, feature selectionmethods, and datasets used in the field.

Prasad, S., & Pal, S. (2020). Hybrid spam email detection using machine learning. International
Journal of Advanced Research in Computer Science, 11(4), 131-135. This study proposes a hybrid
approach for email spam detection using machine learning algorithms. It combines the strengths of
multiple classifiers to improve overall classification accuracy.

Kaur, G., & Kaur, M. (2020). Review on email spam detection using machine learning techniques.
In Proceedings of the 10th International Conference on CloudComputing, Data Science &
Engineering (pp. 192- 198). This paper provides a comprehensive review of email spam detection
techniques using machine learning. It discusses various algorithms, feature selection methods, and
performance evaluation metrics.
Bharti, S. K., Singh, S., & Malhotra, A. (2019). Machine learning-based spam email detection using
optimized features. In Proceedings of the International Conference on Advanced Computing and
Intelligent Engineering (pp. 147-158). Springer, Singapore. This research proposes a machine
learning-based approach for email spam detection using optimized features. The study evaluates the
performance of different classifiers and feature selection techniques.

Li, Y., He, L., Guo, H., Liu, L., & Zhao, Y. (2019). Deep learning for email spam detection: A
comparative analysis. Neural Computing and Applications, 31(11), 8205-8216. This study compares
different deep learning architectures for email spam detection. It provides insights into the
performance of convolutional neural networks (CNN) and recurrent neural networks (RNN) in this
context.

3.2 SUMMARY
he provided studies offer valuable insights into the field of email spam detection using machine
learning. They explore various techniques, including deep learning architectures, machine learning
algorithms, feature selection methods, and performance evaluation metrics.

Li et al. (2019) compare CNNs and RNNs for spam detection, finding that CNNs generally
outperform RNNs in this task. Bharti et al. (2019) propose a machine learning approach using
optimized features and evaluate different classifiers and feature selection techniques. Kaur and Kaur
(2020) provide a comprehensive review of spam detection techniques, covering algorithms, feature
selection, and evaluation metrics. Prasad and Pal (2020) propose a hybrid approach combining
multiple classifiers for improved accuracy.

Mishra et al. (2020) present a comprehensive review of spam detection techniques, covering
algorithms, feature selection, and datasets. Salam et al. (2020) conduct a comparative analysis of
Naive Bayes, SVM, and decision trees for spam detection. Saini and Kumar (2020) compare Naive
Bayes, SVM, and random forests for spam detection. Ahirwal and Kaushik (2021) provide a
comprehensive review of spam detection techniques, covering algorithms, feature extraction, and
evaluation metrics. Rani and Nasa (2021) compare Naive Bayes, decision trees, and K-nearest
neighbors for spam detection.
 CHAPTER 4
Software Requirement
Specification
4.1 INTRODUCTION TO SYSTEM ENGINEERING

4.2 SYSTEM REQUIREMENTS

Hardware Requirements:-
1. Processor: Intel i3/i5/i7
2. Speed: 1.1 GHz
3. RAM: 8 GB (min)
4. Hard Disk: 40 GB
5. Keyboard, Mouse

Software Requirement:-
1. Operating System: Windows, macOS, or Linux
2. Front End: Python
3. Framework: TensorFlow, PyTorch, Keras, and MXNet
4. Libraries: NumPy, Pandas, Scikit-learn, and Matplotlib
4.3 Functional Requirements
Functional requirements describe the specific behavior and functionality that a system
or program must exhibit in order to meet its intended purpose. Based on the code
snippet provided, here are some possible functional requirements for the program:

1. Large Dataset: A substantial amount of labeled email data is required to train the neural
network effectively.
2. Data Quality: The dataset should be clean and free from errors to ensure accurate training
and evaluation..
3. Data Diversity: The dataset should include a variety of spam and non-spam emails to
represent real-world scenarios.
4. Network Depth and Width: The number of layers and neurons per layer should be
carefully chosen to balance accuracy and computational efficiency.
5. Activation Functions: Appropriate activation functions (e.g., ReLU, sigmoid, tanh)
should be selected to ensure effective learning.
6. Weight Initialization: Proper weight initialization techniques (e.g., Xavier initialization,
He initialization) should be used to prevent vanishing or exploding gradients.
7. Loss Function: A suitable loss function (e.g., cross-entropy, mean squared error) should
be chosen to measure the model's performance.
8. Optimization Algorithm: An effective optimization algorithm (e.g., stochastic gradient
descent, Adam) should be used to update the network's weights.
9. Regularization: Techniques like L1 or L2 regularization can help prevent overfitting and
improve generalization.
10. Metrics: Relevant metrics (e.g., accuracy, precision, recall, F1-score) should be used to
evaluate the model's performance.
11. Cross-Validation: Cross-validation techniques (e.g., k-fold cross-validation) should be
employed to assess the model's generalization ability.
12. Real-time Processing: The model should be able to process emails in real-time to prevent
spam from reaching users' inboxes.
13. Scalability: The system should be scalable to handle large volumes of email traffic.
14. Continuous Improvement: The model should be regularly updated with new data and
refined to adapt to evolving spam techniques.
 CHAPTER 5
Results & Discussion
 5.1 RESULTS (SCREEN-SHOTS OF THE RESULT)

5.1.1 Interface Design

To run the project file you need to open the Jupyter Notebook prompt andchange the directory
to the folder where the projects files are present as shown in below figure:

After changing the directory, you need to open the file in below figure:
Click on kernel and select restart and run all.

Wait for some time until the code gets execute, now at prediction template enterthe string
which you want to predict whether it is a spam or ham and click on run as shown below:

If the message is ham it will show Ham as shown below:

 CHAPTER 6
CONCLUSION
6.1 Conclusion
In this study, we reviewed machine learning approaches and their application to the field of spam
filtering. A review of the Neutral neywork algorithms been applied for classification of messages as
either spam or ham is provided. The attempts made by different researchers to solving the problem
of spam through the use of machine learning classifiers was discussed. The evolution of spam
messages over the years to evade filters was examined. The basic architecture of email spam filter
and the processes involved in filtering spam emails were looked into. The paper surveyed some of
the publicly available datasets and performance metrics that can be used to measure the effectiveness
of any spam filter. The challenges of the machine learning algorithms in efficiently handling the
menace of spam was pointed out and comparative studies of the machine learning technics available
in literature was done. We also revealed some open research problems associated with spam filters.
In general, the figure and volume of literature we reviewed shows that significant progress have been
made and will still be made in this field. Having discussed the open problems in spam filtering, further
research to enhance the effectiveness of spam filters need to be done. This will make the development
of spam filters to continue to be an active research field for academician and industry practitioners
researching machine learning techniques for effective spam filtering. Our hope is that research
students will use this paper as a spring board for doing qualitative research in spam filtering using
machine learning, deep leaning and deep adversarial learning algorithms.
6.2 Future Scope
1. Deep Learning Architectures: Explore more advanced deep learning architectures, such as
recurrent neural networks (RNNs) with long short-term memory (LSTM) units or transformer
models, to capture the temporal dependencies and contextual information in email messages.
These architectures could improve the detection of sophisticated spam techniques that involve
sequences of actions or hidden meanings.Enhanced Integration Capabilities, expanding the
IDE's ability to integrate with various third-party services and APIs will improve its versatility
and support a broader range of applications, making it more adaptable to different user
requirements.
2. Hybrid Approaches: Combine neural networks with other machine learning techniques, such
as rule-based systems or ensemble methods, to create hybrid models that leverage the
strengths of different approaches. This could lead to more robust and accurate spam detection
systems.

3. Generative Models: Utilize generative adversarial networks (GANs) to generate synthetic

spam examples, expanding the training data and improving the model's ability to detect novel
spam variants. GANs could also be used to create adversarial attacks to test the robustness of
spam detection systems.

4. Contextual Understanding: Develop neural network models that can understand the context
of email messages, considering factors like sender reputation, recipient relationships, and
subject matter. This could help identify spam that is more likely to be missed by traditional
techniques based solely on textual features.
6.3 References
[1] Suryawanshi, Shubhangi & Goswami, Anurag & Patil, Pramod. (2020). Email Spam
Detection: An Empirical Comparative Study of Different ML and Ensemble Classifiers.
69-74. 10.1109/IACC48062.2019.8971582.

[2] Karim, A., Azam, S., Shanmugam, B., Krishnan, K., & Alazab, M. (2020). A
Comprehensive Survey for Intelligent Spam Email Detection. IEEE Access, 7, 168261-
168295. [08907831]. https://doi.org/10.1109/ACCESS.2019.2954791

[3] K. Agarwal and T. Kumar, "Email Spam Detection Using Integrated Approach of Naïve
Bayes and Particle Swarm Optimization," 2018 Second International Conference on
Intelligent Computing and Control Systems (ICICCS), Madurai, India, 2018, pp. 685-690.

[4] Harisinghaney, Anirudh, Aman Dixit, Saurabh Gupta, and Anuja Arora. "Text and image-
based spam email classification using KNN, Naïve Bayes and Reverse DBSCAN
algorithm." In Optimization, Reliabilty, and Information Technology (ICROIT), 2014
International Conference on, pp.153-155. IEEE, 202

[5] Carreras, X., & Marquez, L. (2001). Boosting trees for anti-spam email filtering.
Proceedings of the Conference on Recent Advances in Natural Language Processing, 9-
15.