Beyond Skill Gap Analysis: An Explainable Transformer-Based Framework for Software Workforce Intelligence

Abstract

The rapid evolution of the software industry has created a persistent skill gap between academic curricula and industry requirements. Traditional approaches to skill gap analysis-which rely on manual comparison of course outlines with job descriptions-are time-consuming and fail to capture dynamic market trends. This paper presents Beyond Skill Gap Analysis, a novel explainable transformer-based framework for dynamic software workforce intelligence. The proposed system uses BERTbased models to extract both technical and soft skills from unstructured job postings with high accuracy, achieving an F1-score of 0.89-significantly outperforming traditional keyword-based methods $(\mathrm{F}1=0.76)$. The framework incorporates explainable AI techniques (LIME and SHAP) to provide transparency in the decision-making process, enabling stakeholders to understand the rationale behind skill identification. Additionally, LSTM-based time series forecasting enables proactive workforce planning by predicting emerging skill trends. Experimental results demonstrate that the system achieves 90% accuracy in skill identification, with precision of 0.91 and recall of 0.87. The findings provide educational institutions, organizations, and policymakers with actionable insights to bridge the skill gap in the software industry and prepare a future-ready workforce.