ANALISIS KOMPARATIF KLASIFIKASI SENTIMEN PENGGUNA APLIKASI INVESTASI MENGGUNAKAN ALGORITMA HYBRID CNN-LSTM, CNN-GRU DENGAN IMPLEMENTASI SMOTE

Abstract

Sentiment analysis, a branch of Natural Language Processing (NLP), plays a crucial role in identifying and classifying opinions embedded in text. This study aims to compare the performance of hybrid CNN-LSTM and CNN-GRU models in sentiment analysis of user reviews for investment applications on Google Play Store, utilizing the Synthetic Minority Over-sampling Technique (SMOTE) to address data imbalance. A total of 15,000 user reviews were collected through web scraping, preprocessed using the TF-IDF method and various text cleaning techniques. The CNN-LSTM and CNN-GRU models were evaluated using an 80%-20% train-test split. The evaluation results showed that CNN-GRU outperformed in terms of precision (91.62%), F1 score (90.45%), and overall accuracy (87.60%), while CNN-LSTM excelled in recall (91.08%) for detecting positive reviews. CNN-GRU was deemed more balanced in detecting both positive and negative sentiments, making it a more reliable choice for sentiment analysis requiring uniform performance