A Review of System for Recommending Jobs Using Machine Learning

The purpose of this research is to conduct a literature assessment of job recommender systems (JRS) that have been published in the recent past. When compared to our prior evaluations of the relevant literature, we placed a greater amount of importance on contributions that took into account the temporal and reciprocal aspects of job recommendations. Previous research on JRS has shown that it may be possible to enhance model performance by taking different perspectives like these into consideration when designing JRS. Additionally, it may result in a more balanced distribution of applicants among a group of occupations that are comparable to one another. In addition to this, we look at the literature from the point of view of the fairness of algorithms. When we looked into this, we discovered that this topic is seldom brought up in the academic literature, and when it does, many writers make the incorrect assumption that deleting the discriminating characteristic would be enough. When referring to the kinds of models that are used in JRS, writers usually refer to their approach as being “hybrid.” In doing so, however, they unfortunately conceal what exactly these procedures include. We divided this expansive class of hybrids into more manageable subclasses by making use of the recommender taxonomies that were already in existence. In addition, we come to the conclusion that the availability of data, and more specifically the availability of click data, has a significant bearing on the selection of a validation technique. Last but not least, despite the fact that the generalizability of JRS across various datasets is seldom taken into consideration, the findings imply that error scores may change across these datasets.