Home
Login.
Artikelilmiahs
50225
Update
RAFIF IMADUDDIN YUDONO
NIM
Judul Artikel
ANALISIS SENTIMEN ULASAN PRODUK E-COMMERCE MENGGUNAKAN MODEL LONG SHORT TERM MEMORY
Abstrak (Bhs. Indonesia)
Penelitian ini membandingkan kinerja model Recurrent Neural Network (RNN) dan Long Short-Term Memory (LSTM) dalam mengklasifikasikan sentimen ulasan produk e-commerce (positif, negatif, netral) menggunakan dataset publik “171k Product Review with Sentiment” dari Kaggle. Setelah tahap pra-pemrosesan teks (pembersihan simbol, angka, stopwords, penanganan data kosong) dan penyeimbangan kelas dengan SMOTE, data diubah melalui tokenization dan padding sebelum dilatih pada kedua model. Hasil evaluasi menunjukkan LSTM unggul dengan akurasi 85% dibandingkan RNN 84%, terutama dalam menangani kelas minoritas (netral) dengan precision lebih tinggi dan ketahanan lebih baik terhadap overfitting, sehingga dinilai lebih efektif untuk analisis sentimen ulasan produk e-commerce.
Abtrak (Bhs. Inggris)
This literature compares the performance of the Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) models in classifying product review sentiments on e-commerce platforms (positive, negative, neutral) using the public dataset “171k Product Review with Sentiment” from Kaggle. After the text preprocessing stage (removal of symbols, numbers, and stopwords, as well as handling missing data) and class balancing using SMOTE, the data was transformed through tokenization and padding before being trained on both models. The evaluation results show that LSTM outperforms RNN with an accuracy of 85% compared to RNN’s 84%, particularly in handling minority classes (neutral) with higher precision and better resistance to overfitting, making it more effective for sentiment analysis of e-commerce product reviews.
Kata kunci
Pembimbing 1
Pembimbing 2
Pembimbing 3
Tahun
Jumlah Halaman
Save