SMS Spam and Scam Message Detection Using Machine Learning (Project Guide)

By EngrHenryTech Publishing · Computer Science

SMS SPAM AND SCAM MESSAGE DETECTION USING MACHINE LEARNING: A COMPLETE PROJECT GUIDE

Departments: Computer Science, Cyber Security, Information Technology, Software Engineering, Data Science, Computer Engineering

ABOUT THIS GUIDE

This is a detailed project guide and framework for a machine learning project, not a finished project or ready-made code. It gives you a ready-to-adapt Chapter One, a map for your literature review, a full methodology (data, preprocessing, feature extraction, model training and evaluation), an implementation plan in Python, templates for reporting your results, and guidance for writing your documentation. You obtain the data, write and run the code, and report the results you actually get.

SECTION 1: TOPIC AND ALTERNATIVE TITLES

Main topic:
SMS Spam Detection Using Machine Learning: A Comparative Study of Naive Bayes, Logistic Regression and Support Vector Machine Classifiers

Alternative titles:
1. Detection of Fraudulent SMS Messages in Nigeria Using Machine Learning.
2. A Machine Learning Approach to Phishing Email Detection.
3. Comparative Analysis of Text Classification Algorithms for Spam Filtering.
4. Design of a Mobile App that Flags Suspicious SMS Messages Using a Trained Classifier.
5. Fake News Headline Detection Using Machine Learning.
6. Detection of Scam Messages in Nigerian Pidgin and English Using Machine Learning.

SECTION 2: MODEL CHAPTER ONE

1.1 Background to the Study (model text, rewrite in your own words)

Short Message Service (SMS) remains widely used in Nigeria for banking alerts, one-time passwords, notifications and personal communication. It is also widely abused. Users receive unsolicited advertising messages (spam) and fraudulent messages designed to steal money or information, such as fake bank alerts, false prize notifications and messages asking users to "update" their BVN through a link. Such messages waste users' time, cause financial losses and reduce trust in legitimate messages.

Rule-based filters, which block messages containing particular words or sent from particular numbers, are easy for spammers to evade by changing wording or numbers. Machine learning offers an alternative: a classifier learns patterns that distinguish spam from legitimate messages ("ham") from a labelled dataset, and can then classify new messages. Text classification algorithms, such as Naive Bayes, logistic regression and support vector machines (SVM), have been widely used for spam filtering. A public dataset, the SMS Spam Collection (Almeida, Gómez Hidalgo and Yamakami, 2011), containing 5,574 English SMS messages labelled as spam or ham, is commonly used as a benchmark.

This project develops and compares machine learning models for SMS spam detection, evaluates them with appropriate metrics, and [optionally] tests them on a small set of locally collected messages to examine how well they generalise to Nigerian scam messages.

1.2 Statement of the Problem (model)

Nigerian mobile users receive large numbers of spam and scam messages, and many fall victim to fraud. Simple keyword filters are easily evaded. Models trained on public datasets may not perform well on local messages, which often refer to local banks, the BVN, local currency and Nigerian Pidgin. This project addresses the need for an effective classifier and examines its performance on locally relevant messages.

1.3 Aim and Objectives

The aim is to develop and evaluate machine learning models for SMS spam detection. The objectives are to:
1. obtain and prepare a labelled SMS dataset;
2. extract text features using bag-of-words and TF-IDF;
3. train Naive Bayes, logistic regression and SVM classifiers;
4. evaluate the models using precision, recall, F1-score and the confusion matrix;
5. [optional] evaluate the best model on a small set of locally collected, anonymised messages;
6. [optional] build a simple web or mobile interface that classifies a message entered by the user.

WHAT THE FULL GUIDE CONTAINS

The complete guide continues with:
- research questions, significance, scope, limitations and definitions of terms (precision, recall, F1-score, TF-IDF);
- a literature review map: approaches to spam filtering, text preprocessing, feature extraction, classifiers and evaluation, with a plan for reviewing related works;
- a full methodology: the public SMS Spam Collection dataset, optional consented and anonymised local messages, preprocessing, stratified splitting, avoiding data leakage, model tuning and evaluation;
- an implementation plan in Python with scikit-learn, including a code outline and templates for reporting your results;
- a Chapter Four and Five guide, real starting references, and a checklist of common supervisor corrections.

Please note: this is a research guide and framework to help you plan and write your own project. It does not contain invented data or results. You obtain the data, write and run the code, and report the results you actually get.