Apri Fitri Hartono
Universitas Bina Nusantara

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

Hybrid Recommendation System to Determine Promotion Mechanism in Health and Beauty Retail Apri Fitri Hartono; Amalia Zahra
Glosains: Jurnal Sains Global Indonesia Vol. 7 No. 3 (2026): Glosains: Jurnal Sains Global Indonesia
Publisher : Sekolah Tinggi Agama Islam Kuningan

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.59784/glosains.v7i3.802

Abstract

Background: Health and beauty retail companies face challenges in determining effective promotional mechanisms to increase sales and reduce ineffective promotions. Manual promotion decisions risk error because they require complex analysis of historical sales and promotion data. Objective: This study aims to develop and compare the performance of a machine learning-based recommendation system using the Collaborative Filtering (CF), Content-Based Filtering (CBF), Category-weight Hybrid Recommendation (CWHR), and Subcategory-weight Hybrid Recommendation (SWHR) approaches in determining the promotion mechanism in the retail health and beauty category. Methods: This research uses a comparative experimental method evaluating several machine learning-based recommendation models (CF, CBF, CWHR, and SWHR) using historical retail transaction data. The dataset used is historical sales promotion data for the 2022-2024 period of 1,656,232 transaction data consisting of sales data, product categories, subcategories, sales amounts, and types of promotions. The model evaluation was carried out using the Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE) metrics. Results:  Results showed CWHR achieved the best overall performance (lowest RMSE 0.5096, MAE 0.4078), while SWHR achieved the lowest MAPE (20.38%) despite a considerably higher RMSE (1.8566), indicating that the hybrid approach improves prediction accuracy over single CF and CBF models, with the preferred model depending on the metric prioritized. Conclusion: Thus, a hybrid recommendation system can help category managers in determining a more effective and data-driven promotion strategy. The CWHR/SWHR choice depends on whether error magnitude or percentage accuracy is prioritized.