PERBANDINGAN METODE NAÏVE BAYES, SUPPORT VECTOR MACHINE DAN RECURRENT NEURAL NETWORK PADA ANALISIS SENTIMEN ULASAN PRODUK E-COMMERCE

Abstract

Sentiment analysis serves as a valuable tool for capturing consumer opinions and broader public sentiment. Product reviews posted on e-commerce platforms significantly influence product sales. The objective of this research is to perform sentiment analysis on e-commerce product reviews utilizing Naïve Bayes, Support Vector Machine (SVM), and Recurrent Neural Network (RNN) algorithms. The study encompasses data selection, preprocessing, transformation, data mining, and evaluation/interpretation as crucial stages. Moreover, addressing the issue of imbalanced data, particularly the disparity between positive and negative sentiments, is achieved through the application of oversampling techniques utilizing the SMOTE library. This research aims to enhance the understanding of sentiment analysis, its significance in comprehending consumer opinions, and its role in improving product purchase decisions. The sentiment analysis of e-commerce product reviews was conducted using Naïve Bayes, SVM, and RNN algorithms. The opinions were classified as positive, negative, or neutral. Notably, there is a distinction in the data distribution between positive and negative sentiments (imbalanced data), which necessitates distinct treatment within the models. The findings revealed an accuracy of 86% for Naïve Bayes, 88% for SVM, and 96% for RNN with an epoch of 10

Keywords

• analisis sentimen
• naïve bayes
• support vector machine
• recurrent neural network
• smote