Machine learning for employability prediction: A review of techniques and trends in the higher education sector

Abstract

The widening disjunction between academics and industry has become a major issue in the higher education sector and the disjunction has become a cause of concern in employability of graduates globally. Although the number of educational initiatives and skill-related courses has been increasing, a considerable percentage of graduates are still unprepared to meet the demands of the dynamic job market, resulting in their unemployment and lack of skills. The issue under discussion in real life is only enhanced by the fast-changing nature of demands within the industry, absence of individualized career advice, and the lack of incorporation of data-driven decision-making in schools.
To address this question, machine learning (ML) methods have received growing interest in forecasting the results of employability with the help of multidimensional student data, such as academic performance, technical skills, and behavioral traits. Current research has used different algorithms including Support Vector Machines, Decision Trees, Random Forests and ensemble-based methods to predict students into employable or non-employable with promising predictive accuracy. Nonetheless, these methods are mostly concerned with the predictive net whereas not much is addressed on the actionable insights, skill improvement techniques, and practical use. Moreover, variations in the choice of features, lack of standard datasets, and limited longitudinal validation hamper the generalizability and effect of the existing solutions.
These constraints underscore the necessity to have a deeper insight into the ways predictive analytics can go beyond being classified into a skillful use of employability to mean improvement. It is urgently needed that the existing approaches be methodically reviewed, upcoming trends revealed, and that important gaps in research are determined that prevent effective application of the ML-based employability solutions within higher education.
This paper satisfies this requirement by proposing an organization and systematic review of machine learning methods applied in predicting employability in the higher education setting. It combines the works of other authors to both examine widely used algorithms, feature sets, evaluation measures, and problem domains as well as discuss important trends, including the movement towards ensemble learning and data-driven career support systems. More to the point, this research can identify some serious gaps in the studies, specifically, the inability to combine frameworks of skills development, narrow down on individualized recommendations, and a lack of real-world validation.
The originality of the current paper is that its synthesis of the field is both thorough and critical rather than descriptive as it seeks to offer gap-driven view of what connects predictive analytics to the enhancement of employability. It offers a conceptual path of future research providing an emphasis on the shift of prediction-focused models to intervention-based systems that proactively help to develop skills and prepare careers.
This study is likely to be beneficial to a variety of stakeholders, such as academic institutions that want to enhance the design of their curricula and student outcomes, policymakers wishing to decrease the education-employment gap, researchers interested in pioneering applications of highly advanced ML in the educational field, and students who might eventually have access to more personalized and data-driven career guidance technologies.