Gradient Boosted Email Classification through Integration of Co-Occurrence Network Features and Knowledge-Enhanced Semantics

Modeling and Simulation, Год журнала: 2025, Номер 14(03), С. 222 - 237

Опубликована: Янв. 1, 2025

Язык: Английский

---

BaitBlock: Hybrid AI-Approach for Phishing Detection Across Communication Platforms

Lecture notes in computer science, Год журнала: 2025, Номер unknown, С. 18 - 37

Опубликована: Янв. 1, 2025

Язык: Английский

---

EGMA: Ensemble Learning-Based Hybrid Model Approach for Spam Detection

Applied Sciences, Год журнала: 2024, Номер 14(21), С. 9669 - 9669

Опубликована: Окт. 23, 2024

Spam messages have emerged as a significant issue in digital communication, adversely affecting users’ mental health, personal safety, and network resources. Traditional spam detection methods often suffer from low rates high false positives, underscoring the need for more effective solutions. This paper proposes EGMA model, an ensemble learning-based hybrid approach SMS messages, which integrates gated recurrent unit (GRU), multilayer perceptron (MLP), autoencoder models utilizing majority voting algorithm. The model enhances performance by incorporating additional statistical features extracted message content employing text vectorization techniques, such Term Frequency–Inverse Document Frequency (TF-IDF) CountVectorizer. proposed achieved impressive classification accuracies of 99.28% on Collection dataset, 99.24% Email 99.00% Enron-Spam 98.71% Super 95.09% UtkMl’s Twitter dataset. These results demonstrate that outperforms individual existing literature, providing robust solution enhancing effectively mitigating threats pose communication.

Язык: Английский

---

Advanced Analysis of Learning-based Spam Email Filtering Methods Based on Feature Distribution Differences of Dataset

IEEE Access, Год журнала: 2024, Номер 12, С. 167313 - 167323

Опубликована: Янв. 1, 2024

Язык: Английский

---

Leveraging Large Language Models in Tourism: A Comparative Study of the Latest GPT Omni Models and BERT NLP for Customer Review Classification and Sentiment Analysis

Information, Год журнала: 2024, Номер 15(12), С. 792 - 792

Опубликована: Дек. 10, 2024

In today’s rapidly evolving digital landscape, customer reviews play a crucial role in shaping the reputation and success of hotels. Accurately analyzing classifying sentiment these offers valuable insights into satisfaction, enabling businesses to gain competitive edge. This study undertakes comparative analysis traditional natural language processing (NLP) models, such as BERT advanced large models (LLMs), specifically GPT-4 omni GPT-4o mini, both pre- post-fine-tuning with few-shot learning. By leveraging an extensive dataset hotel reviews, we evaluate effectiveness predicting star ratings based on review content. The findings demonstrate that family significantly outperforms model, achieving accuracy 67%, compared BERT’s 60.6%. GPT-4o, particular, excelled contextual understanding, showcasing superiority LLMs over NLP methods. research underscores potential using sophisticated evaluation systems hospitality industry positions transformative tool for analysis. It marks new era automating interpreting feedback unprecedented precision.

Язык: Английский

---

Comparative Investigation of Traditional Machine-Learning Models and Transformer Models for Phishing Email Detection

Electronics, Год журнала: 2024, Номер 13(24), С. 4877 - 4877

Опубликована: Дек. 11, 2024

Phishing emails pose a significant threat to cybersecurity worldwide. There are already tools that mitigate the impact of these by filtering them, but only as reliable their ability detect new formats and techniques for creating phishing emails. In this paper, we investigated how traditional models transformer work on classification task identifying if an email is or not. We realized models, in particular distilBERT, BERT, roBERTa, had significantly higher performance compared like Logistic Regression, Random Forest, Support Vector Machine, Naive Bayes. The process consisted using large robust dataset applying preprocessing optimization maximize best result possible. roBERTa showed outstanding capacity identify achieving maximum accuracy 0.9943. Even though they were still successful, performed marginally worse; SVM best, with 0.9876. results emphasize value sophisticated text-processing methods potential improve security thwarting attempts.

Язык: Английский

---