IMPLEMENTASI ENSEMBLE LEARNING UNTUK ANALISIS SENTIMEN PADA APLIKASI REKU

AISYAH, RATIH (2025) IMPLEMENTASI ENSEMBLE LEARNING UNTUK ANALISIS SENTIMEN PADA APLIKASI REKU. Undergraduate thesis, UPN "VETERAN" JAWA TIMUR.

	Text (cover) 20082010190-cover.pdf Download (806kB)
	Text (bab 1) 20082010190-bab1.pdf Download (385kB)
	Text (bab 2) 20082010190-bab2.pdf Restricted to Repository staff only until 20 June 2027. Download (645kB)
	Text (bab 3) 20082010190-bab3.pdf Restricted to Repository staff only until 20 June 2027. Download (614kB)
	Text (bab 4) 20082010190-bab4.pdf Restricted to Repository staff only until 20 June 2027. Download (1MB)
	Text (bab 5) 20082010190-bab5.pdf Download (215kB)
	Text (daftar pustaka) 20082010190-daftarpustaka.pdf Download (200kB)
	Text (lampiran) 20082010190-lampiran.pdf Restricted to Repository staff only Download (593kB)

Abstract

The popularity of cryptocurrency in Indonesia has experienced significant growth, as reflected in the increasing number of crypto investors and improving transaction volumes. With the growing public interest in financial applications like Reku, sentiment analysis has become an essential tool to understand user opinions and satisfaction. However, conventional sentiment analysis methods have limitations in processing the complexity of language and meaning in user reviews. This study applies an ensemble learning approach by combining three classification architectures: Convolutional Neural Network (CNN), Support Vector Machine (SVM) with Word2Vec embedding, and the transformer-based model IndoBERT. The performance of each individual model as well as their combined ensemble results is evaluated using soft voting and weighted soft voting methods. The evaluation is carried out using accuracy, precision, recall, and F1-score metrics to measure the impact of model combination on improving sentiment analysis accuracy. Experimental results show that the ensemble learning approach with the weighted soft voting method achieves the best accuracy of 91%, outperforming each individual model. As an implementation, the optimal model is developed into a web-based application using Streamlit, which can be used as a tool to evaluate user opinions on the Reku application. This application is capable of achieving an accuracy of 90% in classifying user sentiments.

Item Type:

Thesis (Undergraduate)

Contributors: