Penerapan Text Mining Pada E-Commerce Analytics Tool Untuk Klasifikasi Penjualan Produk Pangan Olahan Ilegal.

Abstract

E-commerce telah menjadi salah satu sektor ekonomi digital terbesar di Indonesia, dengan kategori makanan dan minuman, termasuk pangan olahan, menempati urutan teratas sebagai barang yang paling banyak dicari pada tahun 2023. Pertumbuhan ini juga meningkatkan risiko peredaran obat dan makanan ilegal secara daring. Badan Pengawas Obat dan Makanan (BPOM) melalui Direktorat Cegah Tangkal, telah mengembangkan E-Commerce Analytics Tool (EAT) guna membantu pencegahan peredaran produk ilegal di situs e-commerce. Penelitian ini bertujuan untuk menemukan metode klasifikasi yang paling tepat dan mengembangkan mesin klasifikasi pada EAT guna mengidentifikasi produk pangan olahan ilegal secara otomatis. Metode klasifikasi yang diuji dalam penelitian ini meliputi Multinomial Na

E-commerce has become one of the largest digital economic sectors in Indonesia. BPS states that food and beverages, including processed food, occupy the top spot in the category of goods most searched for in e-commerce in 2023. This growth can also increase the risk of illegal drug and food distribution online. BPOM developed the E-Commerce Analytics Tool (EAT) to help prevent the circulation of illegal products on e-commerce sites. Analytical tools are needed to identify illegal processed food products. This research aims to find the most appropriate classification method and develop a classification engine in EAT to identify products automatically. Multinomial Naive Bayes, Random Forest, and K-Nearest Neighbors as well as the Bag of Words and TF-IDF feature extraction methods were used as candidate classification methods in this research. Random Forest with TF-IDF outperforms other methods with a precision of 0.860, recall of 0.848, and F1 score of 0.854. This model is then implemented in the form of a model inference workflow to identify illegal processed food products automatically.