ARNAWA, I PUTU THE FLY (2025) HYBRID PREDICTIVE MODEL DALAM PENINGKATAN PREDIKSI KEPUASAN PENGGUNA APLIKASI TRIV. S1 thesis, Universitas Mercu Buana Jakarta.
|
Text (HAL COVER)
01. COVER.pdf Download (363kB) | Preview |
|
![[img]](https://repository.mercubuana.ac.id/style/images/fileicons/text.png) |
Text (BAB I)
02. BAB 1.pdf Restricted to Registered users only Download (240kB) |
|
|
Text (BAB II)
03. BAB 2.pdf Restricted to Registered users only Download (431kB) |
|
|
Text (BAB III)
04. BAB 3.pdf Restricted to Registered users only Download (362kB) |
|
|
Text (BAB IV)
05. BAB 4.pdf Restricted to Registered users only Download (390kB) |
|
|
Text (BAB V)
06. BAB 5.pdf Restricted to Registered users only Download (132kB) |
|
|
Text (DAFTAR PUSTAKA)
07. DAFTAR PUSTAKA.pdf Restricted to Registered users only Download (242kB) |
|
|
Text (LAMPIRAN)
08. LAMPIRAN.pdf Restricted to Registered users only Download (246kB) |
Abstract
This study aims to improve the accuracy of sentiment classification of user reviews for the Triv application through a comparative and hybrid approach. In the initial phase, a performance comparison was conducted between the Support Vector Machine (SVM) and Random Forest algorithms using a dataset of 3,000 reviews collected from the Google Play Store. The data were obtained through purposive sampling, and sentiment classification was based on user rating scores. The results showed that Random Forest exhibited more stable performance, achieving an accuracy of 72.99%, outperforming SVM. Subsequently, Random Forest was employed for time-based sentiment prediction on reviews from the most recent month, achieving an increased accuracy of 84.42%. To further enhance performance, a hybrid model was developed by combining Random Forest and K-Nearest Neighbors (KNN) using a soft voting scheme. Final evaluation results demonstrated that the hybrid model achieved an accuracy of 85.43% and a weighted F1-score of 0.82. Daily trend visualizations indicated that the model could effectively capture the dynamics of user perceptions in real time. These findings confirm that the hybrid modeling approach is superior for sentiment classification and is highly relevant for implementing user satisfaction monitoring systems. Keywords: Sentiment Analysis, Random Forest, KNN, Hybrid Model, Triv Penelitian ini bertujuan meningkatkan akurasi klasifikasi sentimen ulasan pengguna aplikasi Triv melalui pendekatan komparatif dan hybrid. Tahap awal dilakukan perbandingan performa algoritma Support Vector Machine (SVM) dan Random Forest pada data berjumlah 3.000 ulasan dari Google Play Store, dengan purposive sampling dan klasifikasi sentimen berdasarkan skor rating. Hasilnya, Random Forest menunjukkan kinerja lebih stabil dengan akurasi 72,99%, mengungguli SVM. Selanjutnya, Random Forest digunakan dalam prediksi sentimen berbasis waktu terhadap data ulasan satu bulan terakhir dan mencatat peningkatan akurasi hingga 84,42%. Untuk meningkatkan performa lebih lanjut, dibangun model hybrid yang menggabungkan Random Forest dan K-Nearest Neighbors (KNN) dalam skema soft voting. Evaluasi akhir menunjukkan bahwa model hybrid mencapai akurasi 85,43% dan F1-score tertimbang 0,82. Visualisasi tren harian memperlihatkan model mampu mengikuti dinamika persepsi pengguna secara real-time. Temuan ini membuktikan bahwa pendekatan hybrid model lebih unggul dalam klasifikasi sentimen dan relevan digunakan untuk sistem monitoring kepuasan pengguna. Kata Kunci: Analisis Sentimen, Random Forest, KNN, Hybrid Model, Triv
Actions (login required)
 |
View Item |
