Klasifikasi Rentang Gaji Lowongan Pekerjaan di Glints Wilayah Jabodetabek Menggunakan Regresi Logistik dan Random Forest Berbasis Web Scraping

Abstract

Digital transformation has reshaped the labor market, with online platforms such as Glints serving as large-scale data repositories that connect job seekers with employers. In the Greater Jakarta (Jabodetabek) region, salary information is a critical factor in career decision-making; however, salary-related information asymmetry remains a major challenge. This study begins with a descriptive analysis of 1,497 job vacancies collected through web scraping techniques to examine salary distributions across locations and employment statuses. The salaries were classified into three categories: low salary (≤ IDR 5 million), medium salary (IDR 5–10 million), and high salary (≥ IDR 10 million). The results indicate that the majority of job vacancies fall into the low-salary category (77.09%), followed by the medium-salary category (21.37%), while high-salary positions constitute only 1.54% of the total dataset. Subsequently, this study aims to develop salary category classification models by comparing two machine learning methods: Logistic Regression and Random Forest. Model performance was evaluated using accuracy, precision, recall, and F1-score under multiple training–testing split scenarios. The experimental results demonstrate that Random Forest consistently outperforms Logistic Regression, achieving a highest accuracy of 98.00%, compared to approximately 79% for Logistic Regression. These findings suggest that the relationship between job characteristics and salary categories is complex and non-linear, making it more effectively captured by ensemble-based, non-linear algorithms such as Random Forest. This study contributes to improving salary transparency and supports the development of more accurate and data-driven salary prediction systems.