Ismail, Ronizam (2025) An enhanced relative proportional difference model to analyze sentiment intensity of Malaysian telecommunication opinions. Doctoral thesis, Universiti Teknikal Malaysia Melaka.
![[img]](http://eprints.utem.edu.my/style/images/fileicons/text.png) |
Text (24 Pages)
An enhanced relative proportional difference model to analyze sentiment intensity of Malaysian telecommunication opinions (24 pages).pdf - Submitted Version Download (636kB) |
|
Text (Full Text)
An enhanced relative proportional difference model to analyze sentiment intensity of Malaysian telecommunication opinions.pdf - Submitted Version Restricted to Registered users only Download (2MB) |
Abstract
Sentiment intensity analysis is an advanced process within sentiment analysis that not only determines the polarity of opinions but also measures the degree of sentiment strength, offering deeper insights into public opinion. In the Malaysian telecommunications domain, where customer feedback is often expressed in English, Malay, or a mixture of both on informal and noisy social media platforms, the widely used Relative Proportional Difference (RPD) model has been applied for proportional sentiment scoring but suffers from instability, inconsistent outputs, and sensitivity to small threshold changes, leading to unreliable sentiment intensity scores. Thus, this study aims to develop an Enhanced RPD model to improve stability, sensitivity, and consistency in scoring by identifying features that enhance sentiment analysis, designing a model that ensures consistent scoring across multilingual datasets, and evaluating its performance against the baseline RPD. Customer feedback was collected from social media platform and pre-processed for sentiment analysis. The enhanced model incorporates a 0.1 smoothing factor into the original formula, mitigating threshold instability while retaining proportional scoring logic. Its performance was compared against the baseline model using accuracy, F1 score, and stability assessment across multilingual data scenarios. Based on English, Malay, and mixed-language telecom customer feedback, the Enhanced RPD model demonstrates an accuracy of 78.6%, representing a 6.5% improvement over the baseline model, and an F1 score of 78.9%, indicating a 5.7% improvement, and reduces Mean Squared Error by 20% from 0.089 to 0.071. The findings suggest that the Enhanced RPD model provides a more reliable and robust approach for sentiment intensity analysis in multilingual and noisy data environments. This contributes to more accurate sentiment-driven decision-making in the telecommunications sector and holds potential for application in other multilingual industries facing similar analytical challenges. These findings confirm that the Enhanced RPD model delivers more stable, accurate, and fine-grained sentiment intensity scores, addressing multilingual sentiment analysis gaps, providing significant and actionable insights for industry, and offering a reliable foundation for future research in complex real-world data environments
| Item Type: | Thesis (Doctoral) |
|---|---|
| Uncontrolled Keywords: | Sentiment analysis, Sentiment intensity, Sentiment strength, Telecommunication, Relative proportional difference, |
| Subjects: | T Technology T Technology > TK Electrical engineering. Electronics Nuclear engineering |
| Divisions: | Faculty of Information and Communication Technology |
| Depositing User: | Norhairol Khalid |
| Date Deposited: | 21 Jan 2026 07:06 |
| Last Modified: | 21 Jan 2026 07:06 |
| URI: | http://eprints.utem.edu.my/id/eprint/29390 |
| Statistic Details: | View Download Statistic |
Actions (login required)
 |
View Item |