Research on Adaptive Decision Support System for Student Career Planning Based on Multi-Source Academic and Behavioural Data Fusion

Abstract

In the current highly dynamic and competitive global labour market, career planning and decision-making among higher education students have become critical factors directly influencing their future personal development and the effective allocation of human resources. Traditional career guidance models often rely heavily on static psychological assessment questionnaires and individual academic performance indicators, making it difficult to capture the multidimensional potential of students comprehensively or respond adaptively to rapidly changing industry requirements. To address these limitations, this study develops and implements a novel Adaptive Decision Support System (ADSS) designed to provide students with scientific, personalised, and dynamically evolving career path recommendations by integrating heterogeneous academic data with students’ behavioural data generated through digital learning platforms. At the system architecture level, the study proposes a two-stage cascading framework that integrates machine learning prediction with fuzzy multi-criteria decision-making. The prediction module employs the eXtreme Gradient Boosting (XGBoost) algorithm to extract high-dimensional features from unstructured Learning Management System (LMS) interaction data and structured academic records, thereby capturing students’ implicit soft skills and explicit hard skills while generating objective matching probabilities for different career options. The decision-making module subsequently incorporates multi-criteria decision-making (MCDM) theory to examine the interrelationships among career selection criteria using the fuzzy decision-making trial and evaluation laboratory (Fuzzy DEMATEL) method and determine their objective weights. Fuzzy Technique for Order Preference by Similarity to Ideal Solution (Fuzzy TOPSIS) is then applied to integrate students’ subjective preferences and cognitive uncertainty, enabling the comprehensive evaluation and optimisation of alternative career pathways. The results of large-scale empirical analysis indicate that incorporating behavioural data increases the system’s career compatibility prediction accuracy to 93.42%, substantially outperforming the traditional single-source data baseline model. Furthermore, the integrated MCDM framework effectively addresses conflicts among multiple objectives and uncertainties in subjective preferences, allowing the resulting career planning recommendations to remain both data-driven and interpretable while accounting for individual value preferences. The study therefore contributes to the interdisciplinary integration of educational data mining (EDM) and fuzzy decision science while offering higher education institutions a practical system-level solution for implementing targeted and intelligent career guidance interventions.