Dalam era digital yang semakin berkembang, ulasan pengguna menjadi salah satu tolak ukur penting dalam mengevaluasi kualitas layanan pada aplikasi e-commerce. Tokopedia, sebagai salah satu marketplace terbesar di Indonesia, menyediakan platform ulasan melalui Google Play Store, yang mencakup opini positif, negatif, atau netral dari para penggunanya. Penelitian ini menggunakan algoritma Naive Bayes, sebuah metode supervised learning yang dikenal efisien untuk pengklasifikasian data teks. Data dikumpulkan melalui proses scraping dari Google Play Store, mencakup 5000 ulasan berbahasa Indonesia dalam periode Mei 2024 hingga Juli 2024. Tahapan penelitian meliputi preprocessing teks, seperti case folding, tokenizing, stopword removal, dan stemming, diikuti dengan pembobotan kata menggunakan metode Term FrequencyInverse Document Frequency (TF-IDF). Model Naive Bayes kemudian diterapkan untuk mengklasifikasikan sentimen ulasan ke dalam tiga kategori: positif, negatif, dan netral. Hasil penelitian menunjukkan bahwa Naive Bayes dapat mengklasifikasikan sentimen ulasan dengan akurasi 73%, mencapai precision, recall, dan F1-score yang tinggi untuk sentimen positif maupun negatif. Temuan ini mendukung efektivitas Naive Bayes dalam analisis sentimen ulasan e-commerce.

In the rapidly evolving digital era, user reviews have become a crucial benchmark for evaluating service quality in e-commerce applications. Tokopedia, as one of the largest marketplaces in Indonesia, provides a review platform via Google Play Store, which includes positive, negative, or neutral opinions from its users. This study utilizes the Naive Bayes algorithm, a supervised learning method known for its efficiency in text classification. Data was collected through a scraping process from the Google Play Store, encompassing 5,000 Indonesian-language reviews from May 2024 to July 2024. The research stages included text preprocessing techniques such as case folding, tokenizing, stopword removal, and stemming, followed by word weighting using the Term Frequency-Inverse Document Frequency (TF-IDF) method. The Naive Bayes model was then applied to classify sentiment into three categories: positive, negative, and neutral. The results showed that Naive Bayes achieved a 73% accuracy in classifying sentiment, with high precision, recall, and F1-score for both positive and negative sentiments. These findings support the effectiveness of the Naive Bayes algorithm in sentiment analysis for e-commerce reviews.