|
The Public Bus Complaint Classification by Deep Learning Model |
|---|---|
| รหัสดีโอไอ | |
| Creator | Chakkarin Santirattanaphakdi |
| Title | The Public Bus Complaint Classification by Deep Learning Model |
| Contributor | Suphakit Niwattanakul |
| Publisher | Department of Information ScienceFaculty of Humanities and Social Sciences, Khon Kaen University |
| Publication Year | 2565 |
| Journal Title | Journal of Information Science |
| Journal Vol. | 40 |
| Journal No. | 1 |
| Page no. | 19-41 |
| Keyword | Text Classification, Machine Learning, Deep Learning, Public Bus |
| URL Website | https://www.tci-thaijo.org/index.php/jiskku/index |
| Website title | Journal of Information Science |
| ISSN | 2773-8841 |
| Abstract | Purpose: The objectives of this study were to design and develop the public bus complaint classification using the deep learning approach and to assess the complaint classification accuracy.Methodology: This investigation began with the design and development of public bus complaint classification using deep learning approach to create a binary model for dividing words by deep-cut algorithm. After that, all words were transformed into a word bag for word indexing which could calculate keyword weights for classifying statements to mitigate the data duplication problem with t-SNE method in order to make distribution of similar words for formulating 7 classes of public bus complaint classification, including driving class, shuttle stop class, service provider class, bus schedule operation class, vehicle class, service equipment supply class, and epidemic prevention class.Findings: The findings revealed that all classes were found to be accurate at 90%. However, the service provider class which received the largest number of the public bus complaints achieved its accuracy only 83%. This resulted in a variety of different information based on the user context. The comparison between the deep-cut algorithm and the fast-text algorithm showed that they both attained the high accuracy in the same direction. When measuring the accuracy of the service provider problem tagging, it was found that the accuracy reached good level. This indicated that the public bus complaint classification with deep-cut learning approach provided accurate results, especially the accuracy of one complaint per one issue due to the suitability of the issue terms which could be defined by the context. On the contrary, many problems found in the study were caused by duplication of terms in some classes. For instance, 'loudness' could be attributed to loud noise from drivers in the driving class or from the engine sound in the vehicle class or in the equipment supply class. Therefore, the term accuracy had to be semantically and operationally measured.Application of the study: The study findings are of benefit to responsible individuals for improving their services to meet their user service needs. |
