Development of a Job Recommendation System on LinkedIn Based on User CV Using Content-Based Filtering Method

Technological advancements have transformed social media into a crucial tool across various aspects of life, including communication, entertainment, and education. In the professional world, platforms like LinkedIn play a vital role in personal branding, recruitment, and career development. However, LinkedIn’s job search feature often falls short in providing relevant and personalized recommendations due to the complexity of available information, varied interpretations of job requirements, and misalignment between job descriptions and user profiles. This study aims to develop an effective job recommendation system on LinkedIn using users’ Curriculum Vitae (CV) through a Content-Based Filtering approach. Three text vectorization methods namely TF-IDF, Word2Vec, and Sentence-BERT were compared, with performance evaluated using the Normalized Discounted Cumulative Gain (NDCG) metric, expert validation, and user testing. The results show that Word2Vec is the most optimal model, achieving the highest NDCG score (0.971), followed by SBERT (0.909) and TF-IDF (0.550). Word2Vec was consistently rated as the most relevant and accurate by both experts and users, while SBERT performed better in aligning with users’ career interests. The recommendation system was implemented as a website, which was considered intuitive and user-friendly by most respondents. This research demonstrates that a CV-based recommendation system using Word2Vec has strong potential to enhance job search relevance on LinkedIn and help users discover career opportunities that align more closely with their skills and interests.