Komparasi Kinerja Model Machine learning Berbasis Metadata Produk untuk Prediksi Popularitas Produk Elektronik pada Marketplace Lazada Indonesia

Abstract

This study explores the use of product metadata to predict the popularity of electronic products in the Lazada Indonesia marketplace. By using machine learning models, including Logistic Regression, Random Forest, and XGBoost, this study shows that simple metadata such as product category, brand, price, and rating are sufficiently informative to build predictive models. The results indicate that although Logistic Regression delivers the lowest performance due to its linear nature, both Random Forest and XGBoost provide significant improvement. XGBoost achieves the best results with the highest accuracy and F1-score, making it the most effective model for predicting product popularity. These findings highlight the complexity of e-commerce data, which requires more flexible models to capture non-linear patterns and interactions among product features. This study contributes to e-commerce management by providing insights into the use of machine learning for inventory management, promotional strategies, and product placement in digital marketplaces