Job recommender systems: a systematic literature review, applications, open issues, and challenges

Abstract

BackgroundThe HR department has undergone substantial transformations, with a focus on improving recruitment, employee satisfaction, and training through technological integration, particularly with Recommender Systems (RSs) and Job Recommender Systems (JRSs). In addition, the adoption of remote working across multiple industries has significantly increased the need for smart technologies in Human Resources (HR) processes over time. The COVID-19 pandemic accelerated this shift, placing even greater emphasis on the utilization of these technologies.ObjectiveOver the past decade, research on JRSs has increased significantly; however, no comprehensive Systematic Literature Review (SLR) has been conducted in this field. To fill this gap, this study conducts an SLR covering relevant RS and JRS studies published in the last 13 years. It aims to critically analyze commonly used methodologies, identify key research trends and gaps, provide a comprehensive overview of existing literature, define application areas in HR processes, classify performance evaluation metrics used, highlight underexplored themes, discuss limitations of current research, and offer recommendations for researchers and HR practitioners for future studies in the JRS field.MethodIn this SLR, a total of 19 survey articles (including both RSs and JRSs) and 57 research articles on JRS solutions, published between 2010 and 2023, were reviewed and categorized based on publication years, publishers, application themes, applied methodologies, RS filtering techniques, utilized tools, and future studies. Moreover, a comprehensive taxonomy was developed to classify the filtering techniques and methodologies used in both RSs and JRSs. The study also categorizes and classifies performance assessment metrics used in both RS and JRS solutions into four categories: Predictive Accuracy, Classification-Based Accuracy, Ranking-Based Accuracy, and Non-Accuracy-Based Metrics.ResultsKey findings reveal that while JRS solutions utilizing Collaborative, Content-Based, Hybrid-Based, and Knowledge-Based Filtering are well-represented in the literature, those applying Demographic-Based, Reciprocal-Based, Popularity-Based, and Non-Personalized Filtering remain underexplored. The methodologies highly employed in JRS research include similarity-based techniques (e.g., TOPSIS, Cosine Similarity), machine learning models (e.g., XGBoost, LSTM), and knowledge-based approaches (e.g., Ontology, Semantic Web). Evaluation metrics vary significantly based on objectives and filtering techniques, with Classification-Based Accuracy Metrics (e.g., precision, recall, F1 score) being the most commonly used. An analysis of historical trends in JRS publications revealed fluctuations in research interest, with a decline in 2011 and notable peaks in 2017 and 2023. The study also identified a lack of JRS solutions addressing broader HR processes beyond candidate-job matching.ConclusionAs a result, JRSs are still under development and need further improvements, particularly in reducing bias and increasing efficiency in HR processes through advanced AI technologies. Over the past 13 years, JRS studies have explored a variety of themes, including recommending suitable team members, identifying training modules based on skill gaps, developing career path planning systems, assessing employee satisfaction, standardizing professional knowledge, analyzing job postings to determine the busiest job fields, and estimating job tenure and turnover intention. However, research on other crucial HR processes, such as workforce planning, personalized JRSs such as recruitment of special groups, employee engagement strategies, and talent retention frameworks, remains limited, highlighting opportunities to explore in the future.