<![CDATA[The growth of the skincare industry in Indonesia has driven the need for a product selection system that can help consumers choose product combinations that suit their skin type and budget. Previous studies generally use Multi-Criteria Decision Making (MCDM) and Machine Learning approaches that only focus on product ranking or match prediction without considering simultaneous multi-product optimization. This study proposes an optimization approach using the Variable Neighborhood Search (VNS) algorithm to determine combinations of four skincare product categories: facial wash, moisturizer, sunscreen, and serum based on skin type suitability and budget constraints. The dataset consists of 173 local skincare products obtained through web scraping from the Sociolla website, followed by data cleaning and extraction of skin type data from product descriptions. The performance of VNS is evaluated by comparing it to Random Search (RS) as a baseline, which is a random selection from a set of feasible solutions that suit skin type and do not exceed budget. In the testing and comparison, four scenarios were designed based on oily and dry skin types with budget variations of Rp150,000 to Rp350,000. Experimental results show that VNS consistently delivers superior performance compared to RS, characterized by a maximum SkinScore of 1, better budget efficiency, and a more stable composite score across all scenarios. Meanwhile, RS produces more variable solutions and is less precise in approaching the budget constraint. This study demonstrates that VNS is an effective and reliable approach for optimizing skincare product combination selection based on consumer needs and cost constraints.]]>
