Published in the Formosa Journal of Science and Technology in 2026, the study combines sentiment analysis, machine learning, automated review collection, and instant notifications into a single operational platform. The system was developed and tested using customer reviews from Aveta Hotel Malioboro, one of Yogyakarta’s well-known hospitality destinations.
The findings suggest that automated sentiment monitoring can help hotels respond faster, protect digital reputation, and convert customer feedback into practical business decisions.
Digital Reputation Has Become a Core Hospitality Asset
For hotels, online reviews increasingly influence customer choices before booking decisions are made. Platforms such as Google Maps function as public reputation channels where positive experiences attract future guests and unresolved negative comments may reduce trust.
The challenge is scale.
By early 2026, Aveta Hotel Malioboro had accumulated more than 3,400 reviews on Google Maps. Monitoring these reviews manually became increasingly inefficient, especially during busy operational periods when hotel staff had limited capacity to continuously check customer feedback.
According to Hamzah and Pramudwiatmoko from Universitas Teknologi Yogyakarta, delayed responses to negative reviews can reduce service responsiveness and weaken a hotel’s competitive position.
Instead of treating reviews as static feedback, the researchers explored whether hotels could automatically detect customer sentiment and receive immediate alerts.
How the System Works
The research used historical review data collected from Google Maps.
A total of 2,449 customer reviews posted between February 2020 and May 2025 were gathered and processed. The dataset included review text, timestamps, ratings, and user information.
To turn large volumes of customer comments into usable information, the researchers built a workflow that:
- Collected review data automatically
- Cleaned and standardized the text
- Converted written comments into analyzable information
- Classified reviews into positive or negative sentiment
- Sent real-time notifications to hotel management
Two machine learning approaches were evaluated:
- Naive Bayes
- Support Vector Machine (SVM)
The researchers also addressed a common issue in review analysis: positive reviews significantly outnumber negative ones.
After filtering and preparing the dataset:
- 2,152 reviews (87.9%) were classified as positive
- 250 reviews (10.2%) were classified as negative
- 47 reviews remained unclassified due to incomplete or ambiguous text
The imbalance required additional training adjustments so the system would remain effective at detecting negative customer experiences.
Support Vector Machine Delivered the Strongest Results
Performance testing compared both algorithms under different training scenarios.
Initial testing showed both approaches performed strongly, but the best overall results came from Support Vector Machine (SVM) after balancing the training data.
Final performance metrics reached:
- Accuracy: 95.40%
- Precision: 95.14%
- Recall: 95.40%
- F1 Score: 95.22%
These results exceeded several earlier sentiment analysis studies referenced by the authors and demonstrated that review classification can reach production-level performance when paired with domain-specific text preparation.
By contrast, Naive Bayes achieved strong results under the original dataset structure but showed reduced performance after balancing techniques were introduced.
As ethically paraphrased from the authors’ discussion, Rinaldi Hamzah and Dr.Eng. Arif Pramudwiatmoko of Universitas Teknologi Yogyakarta concluded that Support Vector Machine becomes more effective when trained on balanced review data because it learns the separation between positive and negative opinions more reliably in complex text environments.
From Research Prototype to Operational Hotel Tool
A distinguishing contribution of this work is that the research did not stop at algorithm comparison.
The best-performing model was integrated into a functioning web-based monitoring system.
The platform includes:
- A real-time dashboard for review summaries
- Sentiment trend visualization
- Historical review records
- Automated review retrieval
- Subscriber management tools
- Instant Telegram notifications for hotel teams
The notification system allows multiple hotel staff members to receive alerts simultaneously without opening a separate dashboard.
The researchers tested 12 system functions, including review retrieval, duplicate detection, classification performance, notification delivery, and dashboard operation.
All tested functions operated successfully.
Why This Research Matters
The study demonstrates how artificial intelligence can support service industries beyond forecasting and automation.
For hospitality businesses, automated sentiment monitoring creates opportunities to:
- Respond faster to customer dissatisfaction
- Detect service issues earlier
- Strengthen reputation management
- Improve operational decision-making
- Reduce manual monitoring workload
The approach may also be adapted for restaurants, tourism services, public services, healthcare institutions, and other sectors that rely heavily on online customer feedback.
The authors recommend future research using larger datasets from multiple hotels and additional review platforms. They also suggest evaluating modern Indonesian-language AI models such as transformer-based architectures to further improve classification performance.
Author Profiles
Source
Hamzah, R., & Pramudwiatmoko, A. (2026). Sentiment Monitoring of Google Maps Reviews Using Naive Bayes and Support Vector Machine at Aveta Hotel Malioboro. Formosa Journal of Science and Technology, Vol. 5, No. 6, 1569–1588.
URL : https://journalfjst.my.id/index.php/fjst
DOI: https://doi.org/10.55927/fjst.v5i6.95
0 Komentar