Multilabel classification of natural disaster news event using support vector machine / Nur Fatini Md Nazir
Nowadays, the ease of accessing online news through social media like Twitter is a concern for all generations. However, the problem is the process of verifying the accuracy of the information is time-consuming. Plus, the increment amount of online news stories makes it difficult to access disaster-...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2023
|
| Subjects: | |
| Online Access: | https://ir.uitm.edu.my/id/eprint/89017/1/89017.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Nowadays, the ease of accessing online news through social media like Twitter is a concern for all generations. However, the problem is the process of verifying the accuracy of the information is time-consuming. Plus, the increment amount of online news stories makes it difficult to access disaster-relevant news promptly. Therefore, to tackle these issues, a solution is proposed: a multilabel classification system using Support Vector Machine (SVM) for natural disaster news event and using TF-IDF for feature extraction method. The chosen methodology for this project is the waterfall model, known for its systematic and linear software development approach. The results demonstrate the promising outcomes, with SVM model achieving 65.2% accuracy with value C=3 as the parameter of the SVM model. Future work aims to expand the system’s capabilities to classify various types of natural disaster news event from social media, implement automatic classification with diverse techniques or models and extend the language support to include Malay. Incorporating techniques for spell-check, grammar correction and language quality assessment will enhance the system’s ability to accurately classify and extract information from natural disaster news event on social media. In summary, the proposed solution employs SVM-based multilabel classification with TF-IDF feature extraction to address challenges in accessing accurate disaster-relevant news from social media. The system showcases high accuracy and presents possibilities for further enhancements and language expansion. |
|---|