|
Document Type |
: |
Thesis |
|
Document Title |
: |
AN ENHANCED SPAM DETECTION MODEL IN ARABIC TWEETS نموذج محسن للكشف عن التغريدات العربية الغير مرغوب فيها في تويتر |
|
Subject |
: |
Faculty of Computing and Information Technology |
|
Document Language |
: |
Arabic |
|
Abstract |
: |
Spam is an activity that impacts the experience of users on the internet. One of the most popular social networks is twitter, where people exchange short text messages about news, politics, life experiences, etc. Twitter has led to an increase in the spread of spam which is used for advertisements, spread malicious, or just irrelevant content which introduces new security issues and waste of resources. Recently, several approaches have been identified in research for identifying spammers. Even though these approaches presented essential contributions to the field in the English language, few until now covered in the Arabic language. Several challenges are facing existing approaches for Arabic spam detection, especially, handling the morphological nature of Arabic and feature extraction. Therefore, we focus on having a robust spam detection model that could detect the advertisements in trending hashtags in Saudi Arabia. Accordingly, this thesis proposes a deep learning model based on Long Short Term Memory (LSTM) an artificial Recurrent Neural Network (RNN) architecture supported by pre-trained word embedding as a modern feature engineering to detect Arabic spam tweets. Our model achieves an F1-score of 99.28%, which outperforms other machine learning algorithms used in Arabic spam detection. |
|
Supervisor |
: |
Dr. Mohammed Basheri |
|
Thesis Type |
: |
Master Thesis |
|
Publishing Year |
: |
1441 AH
2020 AD |
|
Co-Supervisor |
: |
Dr. Manal Kalkatawi |
|
Added Date |
: |
Friday, August 21, 2020 |
|
Researchers
| مرام محمد دوغان | Dogan, Meram Mohammed | Researcher | Master | |
|