Search for collections on Undip Repository

HYBRID MODEL BOOSTING WEIGHTED EXTREME LEARNING MACHINE DAN RANDOM FOREST UNTUK ANALISIS SENTIMEN REVIEW APLIKASI

YOSEPHINE, Nuki Pujiani and Warsito, Budi and Mutiara K.N., Dinar (2025) HYBRID MODEL BOOSTING WEIGHTED EXTREME LEARNING MACHINE DAN RANDOM FOREST UNTUK ANALISIS SENTIMEN REVIEW APLIKASI. Masters thesis, UNIVERSITAS DIPONEGORO.

[thumbnail of 1. COVER-1.pdf] Text
1. COVER-1.pdf

Download (154kB)
[thumbnail of 1. COVER.pdf] Text
1. COVER.pdf
Restricted to Repository staff only

Download (1MB)
[thumbnail of 2. BAB I.pdf] Text
2. BAB I.pdf

Download (337kB)
[thumbnail of 3. BAB II.pdf] Text
3. BAB II.pdf

Download (823kB)
[thumbnail of 4. BAB III.pdf] Text
4. BAB III.pdf
Restricted to Repository staff only

Download (689kB)
[thumbnail of 5. BAB IV.pdf] Text
5. BAB IV.pdf
Restricted to Repository staff only

Download (982kB)
[thumbnail of 6. BAB V.pdf] Text
6. BAB V.pdf
Restricted to Repository staff only

Download (308kB)
[thumbnail of 7. DAFTAR PUSTAKA.pdf] Text
7. DAFTAR PUSTAKA.pdf

Download (312kB)

Abstract

Penelitian ini dilatarbelakangi oleh kebutuhan untuk memahami persepsi pengguna terhadap aplikasi SIAP UNDIP Mahasiswa melalui analisis sentimen, mengingat ulasan pada Google Play Store menunjukkan distribusi sentimen yang tidak seimbang dan dominasi ulasan negatif. Permasalahan utama penelitian ialah bagaimana mengatasi ketidakseimbangan data dan dimensi fitur teks yang tinggi agar model klasifikasi tidak bias terhadap kelas mayoritas. Tujuan penelitian ini adalah mengembangkan model hybrid berbasis BWELM dan Random Forest untuk meningkatkan akurasi dan kestabilan prediksi sentimen. Metode yang digunakan meliputi pengumpulan data melalui web scraping; pre-processing data dengan pelabelan menggunakan IndoBERT dengan validasi manual, text cleansing, case folding, normalisasi, tokenisasi, stopword removal, dan stemming; ekstraksi fitur menggunakan TF-IDF; serta penerapan model hybrid melalui teknik stacking, dimana BWELM berperan sebagai base learner dan Random Forest sebagai meta-classifier. Hasil pengujian menunjukkan bahwa model hybrid mampu meningkatkan performa dibandingkan model tunggal BWELM dengan akurasi 65,63% dan nilai ROC AUC 77,21%, dintingkatkan menjadi 75,78% untuk akurasi dan nilai ROC AUC 81,40%. Penelitian ini menghasilkan kerangka model analisis sentimen berbasis stacking yang efektif untuk mengatasi ketidakseimbangan data. Hasil penelitian ini dapat menjadi referensi dalam pengembangan sistem informasi yang responsif terhadap persepsi pengguna, sekaligus memperkuat kajian penerapan model hybrid machine learning pada data teks berbahasa Indonesia.
Kata Kunci : Analisis Sentimen, BWELM, Random Forest, Stacked Generalization, Machine Learning, Ulasan Aplikasi

This study aims to understand user perceptions of the SIAP UNDIP Student application through sentiment analysis, driven by the uneven sentiment distribution and the dominance of negative reviews on the Google Play Store. The main challenge is addressing data imbalance and high text feature dimensionality to avoid bias toward the majority class. To solve these issues, this research proposes a hybrid sentiment classification model combining Boosted Weighted Extreme Learning Machine (BWELM) as the base learner and Random Forest as the meta-classifier using a stacking approach. The methodology includes web scraping to collect review data, sentiment labeling with IndoBERT validated manually, and comprehensive text preprocessing: cleaning, case folding, normalization, tokenization, stopword removal, and stemming. Feature extraction is performed using TF-IDF before model training. Experimental results show that the hybrid model significantly improves classification performance over the individual BWELM model. Accuracy rises from 65.63% to 75.78%, while the ROC AUC value increases from 77.21% to 81.40%. This study provides an effective framework for stacking-based sentiment analysis to handle imbalanced datasets and high-dimensional text features. The findings are expected to guide the development of information systems that better reflect user perceptions and contribute to advancing hybrid machine learning applications for Indonesian-language text data.
Keyword : Sentiment Analysis, BWELM, Random Forest, Stacked Generalization, Machine Learning, User Reviews

Item Type: Thesis (Masters)
Uncontrolled Keywords: Analisis Sentimen, BWELM, Random Forest, Stacked Generalization, Machine Learning, Ulasan Aplikasi
Subjects: Sciences and Mathemathic
Divisions: Postgraduate Program > Master Program in Information System
Depositing User: ekana listianawati
Date Deposited: 10 Dec 2025 08:45
Last Modified: 10 Dec 2025 08:45
URI: https://eprints2.undip.ac.id/id/eprint/42037

Actions (login required)

View Item View Item