ONYENA, VICTOR (2026) DEVELOPMENT OF A WEB BASED EMAIL SPAM DETECTION SYSTEM USING MACHINE LEARNING TECHNIQUES. Other thesis, Godfrey Okoye University, Enugu Nigeria.
![[img]](https://eprints.gouni.edu.ng/style/images/fileicons/text.png) |
Text
ONYENA VICTOR.pdf Download (2MB) |
Abstract
The exponential growth of electronic mail as a primary mode of communication has been accompanied by a corresponding rise in unsolicited and often malicious messages, commonly referred to as spam, which compromise user productivity, consume network resources, and expose users to phishing and fraud. While traditional spam filters rely on static, rule-based keyword matching, such approaches struggle to adapt to the evolving tactics employed by spammers, resulting in high false-negative rates and reduced filtering accuracy over time. This study addresses the problem by developing spamshield, a machine learning-based email spam detection system capable of automatically and accurately classifying messages as spam or legitimate (ham). The research employed a dataset of 5,572 labelled sms/email messages obtained from the uci machine learning repository, which was preprocessed using natural language processing techniques including tokenization, stop-word removal, and porter stemming, before being transformed into numerical feature vectors using term frequency- inverse document frequency (tf-idf) vectorization. Three supervised classification algorithms, namely multinomial naïve bayes, logistic regression, and a calibrated support vector classifier, were trained and evaluated on the preprocessed dataset, with performance assessed using accuracy, precision, recall, f1-score, and roc-auc metrics. The trained model was subsequently deployed as a web-based application using a flask rest api backend and an interactive html/css/javascript frontend, allowing users to submit messages for real-time classification. Results obtained from the evaluation showed that all three models achieved strong classification performance, with [insert your best model name] recording the highest accuracy of [insert %] and an f1-score of [insert %], outperforming the other models particularly in correctly identifying spam messages despite the class imbalance present in the dataset. These findings demonstrate that machine learning-based approaches, particularly when combined with effective text preprocessing and tf-idf feature representation, offer a robust and scalable alternative to conventional spam filtering techniques. The study contributes a fully functional, deployable spam detection system and provides a foundation for further research into adaptive, real-time spam filtering solutions.
| Item Type: | Thesis (Other) |
|---|---|
| Subjects: | Q Science > Q Science (General) |
| Divisions: | Faculty of Computing And Information Technology (FACIT) |
| Depositing User: | Cynthia Ugwuoti |
| Date Deposited: | 29 Jul 2026 12:15 |
| Last Modified: | 29 Jul 2026 12:15 |
| URI: | http://eprints.gouni.edu.ng/id/eprint/6072 |
Actions (login required)
 |
View Item |
