IMPLEMENTASI DEEP LEARNING UNTUK ANALISIS SENTIMEN PADA DATA X DALAM PREDIKSI TREN KOREAN STYLE

Abstract

This study implements a deep learning model based on Convolutional Neural Network (CNN) for sentiment analysis of Indonesian-language tweets related to the Korean Style trend. A dataset of 7,187 tweets was collected via web crawling using keywords such as “Korean Style”, “K-pop”, and “Korean fashion”. The preprocessing pipeline included case folding, removal of special characters and emojis, stopword elimination, tokenization, and lemmatization. Word2Vec and FastText embeddings were employed for text representation. The CNN model classified tweets into three sentiment categories: positive, negative, and neutral. Evaluation metrics included accuracy, precision, recall, F1-score, and confusion matrix. Results showed 71.26% validation accuracy with the highest F1-score of 0.81 for the neutral class, while negative sentiment classification remained weak due to class imbalance. Word2Vec outperformed FastText in stability. This research contributes to sentiment analysis in Indonesian social media using deep learning and provides insights into public opinion on Korean cultural trends