Question classification for helpdesk support forum using support vector machine and Naïve Bayes algorithm.
The helpdesk support system is now essential in ensuring the journey of support services runs more systematically. One of the elements that contribute to the non-uniformity of the question data in the Helpdesk Support System is the diversity of services and users. Most questions asked in the system...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Penerbit UTM Press
2023
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/108486/1/SharinHazlinHuspi2023_QuestionClassificationforHelpdeskSupportForum.pdf http://eprints.utm.my/108486/ http://dx.doi.org/10.11113/ijic.v13n1.388 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The helpdesk support system is now essential in ensuring the journey of support services runs more systematically. One of the elements that contribute to the non-uniformity of the question data in the Helpdesk Support System is the diversity of services and users. Most questions asked in the system are in various forms and sentence styles but usually offer the same meaning making its hard for automation of the question classification process. This has led to problems such as the tickets being forwarded to the wrong resolver group, causing the ticket transfer process to take longer response. The key findings in the exploration results revealed that tickets with a high number of transfer transactions take longer to complete than tickets compared to no transfer transaction. Thus, this research aims to develop an automated question classification model for the Helpdesk Support System by applying supervised machine learning methods: Naïve Bayes (NB) and Support Vector Machine (SVM). The domain will use a readily available dataset from the IT Unit. The results using these techniques are then evaluated using confusion matrix and classification report evaluation, including precision, recall, and F1-Measure measurement. The outcomes showed that the SVM algorithm and TF-IDF feature extraction outperformed in terms of accuracy score compared to the NB algorithm. It is expected that this study will have a significant impact on the productivity of team technical and system owners in dealing with the increasing number of comments, feedback, and complaints presented by end-users. |
|---|