Explainable Labor Market Intelligence from Online Job Postings: Multi-Frequency Skill Demand Monitoring and Forecasting

<title>Abstract</title> <p>Rapid digital transformation and artificial intelligence diffusion have intensified the need for timely and interpretable labor market intelligence. While online job postings provide rich real-time signals of employer demand, existing studies often emphasize predictive accuracy while offering limited insight into the persistence, volatility, and strategic meaning of skill demand dynamics. To address this limitation, this study proposes an Explainable Labor Market Intelligence framework that transforms large-scale online job postings into actionable workforce planning signals. Using more than 124,000 LinkedIn job postings from December 2023 to April 2024, the proposed framework constructs Weekly Skill Demand Indices and applies Empirical Mode Decomposition to separate short-term fluctuations from structural trends. Multi-Frequency Hurst Analysis identifies persistent skill demand patterns (H &gt; 0.5), while XGBoost forecasting models achieve R2 = 0.386 in predicting aggregate demand changes. SHAP-based explainability reveals that recent demand momentum and short-term trends are the primary drivers of forecasted changes. The framework generates actionable intelligence signals, identifying healthcare (H = 0.937) and management (H = 0.783) as persistent emerging skills, while information technology (H = 0.825) and sales (H = 0.626) show declining trajectories despite persistent behavior. By integrating skill extraction, persistence analysis, explainable forecasting, and intelligence generation, this study bridges the gap between large-scale online job posting analytics and strategic workforce decision-making, offering marketing analysts and HR strategists an interpretable tool for proactive talent management.</p>