Multi-Aspect Sentiment Analysis of Indonesian Hotel Reviews Using Hybrid Classifier Based on SVM, NB, RF, and K-NN

MEWAR, FAZRAH RAHMAWATI (2025) Multi-Aspect Sentiment Analysis of Indonesian Hotel Reviews Using Hybrid Classifier Based on SVM, NB, RF, and K-NN. Tugas Akhir (S1) - thesis, Universitas Bakrie.

	Text (Cover) 00 cover.pdf - Accepted Version Download (1MB)
	Text (BAB I - III) 01 BAB I - BAB III.pdf - Accepted Version Restricted to Registered users only Download (9MB) | Request a copy
	Text (BAB IV) 02 BAB IV.pdf - Accepted Version Restricted to Registered users only Download (5MB) | Request a copy
	Text (BAB V) 03 BAB V.pdf - Accepted Version Restricted to Registered users only Download (472kB) | Request a copy
	Text (DAFTAR PUSTAKA) 04 DAFTAR PUSTAKA.pdf - Accepted Version Download (1MB)
	Text (LAMPIRAN) 05 LAMPIRAN.pdf - Accepted Version Restricted to Registered users only Download (996kB) | Request a copy

Abstract

Multi-aspect sentiment analysis is a crucial task for understanding detailed user opinions on various facets of a product or service. This study aims to develop and evaluate a robust multi aspect sentiment classification model for Indonesian hotel reviews. Four individual machine learning algorithms, namely Support Vector Machine (SVM), Naive Bayes (NB), K-Nearest Neighbors (K-NN), and Random Forest (RF), are implemented and compared. The mod els are trained using three different feature representation techniques: Bag-of-Words (BoW), Term Frequency-Inverse Document Frequency (TF-IDF), and Word2Vec. Furthermore, a Hy brid Classifier using a stacking methodology is proposed to combine the strengths of the in dividual models. The experiments are conducted on the HoASA dataset from the IndoNLU benchmark. The experimental results demonstrate that the proposed hybrid stacking model achieves a peak accuracy of 93.40%, which was obtained when using Bag-of-Words (BoW) and Term Frequency-Inverse Document Frequency (TF-IDF) feature representations. This fig ure surpasses the performance of the best individual classifier, which was SVM with TF-IDF features, recording an accuracy of 93.10%. Interestingly, in the tests using Word2Vec features, the Random Forest model showed slightly superior performance with an accuracy of 86.10%. The conclusion of this study highlights the effectiveness of the hybrid approach, particularly when paired with classic feature representations like BoW and TF-IDF, in improving the accu racy of multi-aspect sentiment classification.

Item Type:	Thesis (Tugas Akhir (S1) - )
Uncontrolled Keywords:	Sentiment Analysis, Multi-Aspect, SVM, Naive Bayes, Random Forest, K-NN, Stacking, TF-IDF, BoW, Word2Vec
Subjects:	Computer Science > Database management Computer Science > Informatics Computer Science > Information analysis Thesis > Thesis (S1)
Divisions:	Fakultas Teknik dan Ilmu Komputer > Program Studi Informatika
Depositing User:	Fazrah Rahmawati Mewar
Date Deposited:	11 Sep 2025 06:43
Last Modified:	11 Sep 2025 06:43
URI:	https://repository.bakrie.ac.id/id/eprint/12332

Actions (login required)

View Item