Integrasi Data Produk Kosmetik BPOM Dan E-Commerce Menggunakan Bidirectional Neural Network

Handoyo, Kevianwillie (2023) Integrasi Data Produk Kosmetik BPOM Dan E-Commerce Menggunakan Bidirectional Neural Network. Other thesis, Institut Teknologi Sepuluh Nopember.

[thumbnail of 05211940000069-Undergraduate_Thesis.pdf] Text
05211940000069-Undergraduate_Thesis.pdf - Accepted Version
Restricted to Repository staff only until 1 October 2025.

Download (2MB) | Request a copy

Abstract

Frekuensi masyarakat untuk berbelanja online pada e-commerce meningkat dari tahun ke tahun. Produk kosmetik merupakan salah satu produk yang banyak diminati dalam belanja online. Disamping kepopuleran produk kosmetik pada e-commerce, terdapat produk kosmetik dengan kandungan berbahaya yang masih ditemukan di pasaran. BPOM menemukan sejumlah produk kosmetik yang mengandung bahan berbahaya seperti merkuri, asam retinoat, dan hidrokinon. Untuk menentukan apakah suatu produk kosmetik aman untuk dikonsumsi atau tidak, dapat dicek apakah terdapat nama produk atau nomor registrasi yang tertera pada wadah produk pada website resmi BPOM. Namun, terdapat kemungkinan terjadi ketidakcocokan nama produk pada e-commerce dan website BPOM. Oleh karena itu, dibutuhkan solusi berupa rekomendasi produk kosmetik aman yang dapat membantu untuk mengetahui produk kosmetik yang aman untuk dikonsumsi pada platform e-commerce. Pada penelitian ini dilakukan percobaan empat algoritma model, yaitu BRNN, SIF, Attention, dan Hybrid. Data yang digunakan didapatkan dengan melakukan web crawling. Sebelum dilakukan integrasi data, dilakukan praproses data, lalu dilakukan blocking untuk mengeliminasi beberapa pasangan data yang tidak cocok. Kemudian data diberikan label serupa dan tidak serupa untuk setiap pasangan data. Metode pelabelan pasangan data dilakukan secara semi-manual dengan menggunakan jaccard index dan fuzzy string matching sebagai nilai kecocokan pasangan data. Hasil pencocokan pasangan data menggunakan jaccard index memberikan nilai tinggi pada suatu pasangan data walaupun pasangan data tersebut tidak cocok, sedangkan hasil pencocokan fuzzy string matching memberikan nilai kecocokan yang sesuai dengan kemiripan pasangan data. Setelah itu dirancang model yang menggunakan empat algoritma yang diuji. Pada penelitian ini digunakan metode cross validation, dimana model dilakukan training sebanyak 10 kali, kemudian dihitung nilai rata-rata hasil training untuk menentukan tingkat performa model. Performa model dievaluasi menggunakan metrik f1-score, precision, dan recall. Hasil pelatihan terbaik dihasilkan oleh model BRNN dengan nilai f1-score 99.699%. Sementara model dengan pos_neg_ratio 3 dengan epoch 20 menghasilkan f1-score terbaik yaitu 99.91%. Kemudian dirancang aplikasi berbasis web yang memuat 3 fitur utama, yaitu tampilan hasil prediksi pencocokan data, fitur pencarian produk e-commerce, dan fitur prediksi data.
================================================================================================================================
The frequency of people shopping online on e-commerce is increasing from year to year. Cosmetic products are one of the products that are in great demand in online shopping. Despite the popularity of cosmetic products in e-commerce, there are cosmetic products with harmful ingredients that are still found in the market. BPOM found several cosmetic products that contain hazardous ingredients such as mercury, retinoic acid, and hydroquinone. To determine whether a cosmetic product is safe for consumption or not, it can be checked whether there is a product name or registration number listed on the product container on the official BPOM website. However, there is a possibility of mismatching product names in e-commerce and the BPOM website. Therefore, a solution is needed in the form of safe cosmetic product recommendations that can help to find cosmetic products that are safe for consumption on e-commerce platforms. In this study, four model algorithms were tested, namely BRNN, SIF, Attention, and Hybrid. The data used is obtained by doing web crawling. Before data integration, data preprocessing is carried out, then blocking is carried out to eliminate several pairs of data that do not match. Then the data is labeled similar and dissimilar for each pair of data. The data pair labeling method is done semi-manually by using the Jaccard index and fuzzy string matching as the data pair match value. The result of matching data pairs using the jaccard index gives a high value to a data pair even though the data pair does not match, while the result of fuzzy string matching gives a match value that matches the similarity of the data pair. After that, a model is designed that uses four tested algorithms. In this research, the cross validation method is used, where the model is trained 10 times, then the average value of the training results is calculated to determine the level of model performance. Model performance was evaluated using f1-score, precision, and recall metrics. The best training results were generated by the BRNN model with an f1-score value of 99.699%. While the model with pos_neg_ratio 3 with epoch 20 produces the best f1-score of 99.91%. Then a web-based application is designed that contains 3 main features, namely the display of data matching prediction results, e-commerce product search features, and data prediction features.

Item Type: Thesis (Other)
Uncontrolled Keywords: Integrasi Data, Produk Kosmetik, BPOM, Data Integration, Cosmetic Products, BPOM, Bidirectional Recurrent Neural Network,
Subjects: Q Science > QA Mathematics > QA76.9D338 Data integration
R Medicine > R Medicine (General) > R858 Deep Learning
Divisions: Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User: Kevianwillie Handoyo
Date Deposited: 25 Jul 2023 06:39
Last Modified: 25 Jul 2023 06:39
URI: http://repository.its.ac.id/id/eprint/99339

Actions (login required)

View Item View Item