Machine Learning Model Development for Adaptive Recruitment Recommendation System Based on Portfolio Analysis and Professional Network

Abstract

The rapid advancement of digital transformation and artificial intelligence has significantly reshaped recruitment processes within organizations. Conventional recruitment systems predominantly rely on curriculum vitae screening and keyword-based matching, which often fail to capture contextual competencies and relational professional evidence. This study proposes the development of an adaptive machine learning–based recruitment recommendation system that integrates professional portfolio analytics and professional network structures within a unified graphbased framework. The proposed approach adopts a Research and Development (R&D) methodology under a data-driven system development paradigm. Candidate data from an existing recruitment system are integrated with external professional data sources, including GitHub and LinkedIn. A heterogeneous graph representation is constructed to model relationships among candidates, skills, projects, and organizations. Graph Neural Networks (GNN) are employed to learn contextual relational embeddings, while a Gradient Boosting Machine (GBM) is utilized for candidate job suitability classification. The proposed framework is designed to enhance objectivity, contextual awareness, and adaptability in recruitment decision-making. By leveraging multi-source digital professional evidence and incorporating an adaptive learning mechanism, the system aims to reduce skills mismatch and improve alignment between candidate competencies and evolving industry requirements. Future work will focus on empirical validation using real-world recruitment datasets and the integration of fairness-aware and explainable AI mechanisms to ensure transparency and ethical compliance.