This Author published in this journals
All Journal AkMen JURNAL ILMIAH
Claim Missing Document
Check
Articles

Found 1 Documents
Search

PEMANFAATAN MACHINE LEARNING UNTUK ANALISIS ULASAN PELANGGAN DAN KINERJA PENJUALAN DI TOKOPEDIA Agustini, Agustini; Akib, Asrijal
AkMen JURNAL ILMIAH Vol. 22 No. 1 (2025): AkMen JURNAL ILMIAH
Publisher : Lembaga Penelitian, Publikasi dan Pengabdian Pada Masyarakat

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.37476/akmen.v22i1.5451

Abstract

This study analyzes Tokopedia customer reviews to link review characteristics with sales indicators and to deliver an operational ranking tool. Using the public file tokopedia-product-reviews-2019.csv (40,476 reviews, 3,662 products, 158 shops, 5 categories), text was normalized, Indonesian stopwords were removed with Sastrawi, types were fixed, and duplicates were pruned. Star ratings were mapped as weak labels for negative, neutral, and positive sentiment. Descriptive analyses covered rating and sentiment distributions, category sales, top and bottom products, extreme reviews, and the 30 most frequent words. A composite Best Product to Sell index was proposed: 0.5·sold_norm + 0.3·avg_rating_norm + 0.2·pos_ratio_norm, with a minimum review threshold and min–max scaling. Findings show a strong positive skew in ratings and sales concentrated in a small set of SKUs and categories (notably sports, electronics, fashion). Negative reviews focus on shipping delays, packaging safety, and order administration, while positive reviews fit to description, quality, fast shipping, and price. The index highlights candidates for scale up and flags items needing diagnosis, offering a reproducible and extensible workflow for marketplace portfolio decisions.