Combined Application of Various Techniques for Personalized Job Recommendation

Abstract

The Personalized Job Recommender System is a subset of the custom recommendation system that provides a solution to the problem of information overload and is widely applied in numerous domains to solve a plethora of problems, such as unemployment and employment churn that we have seen emerging at higher rates in the COVID era. Furthermore, different jobs require divergent skill sets from their candidates to get hired. In this paper, we analyze the similarity techniques for Job Recommendation Systems based on the research done in the field of Job Recommendations. In our implementation, we have used three similarity measures: Tanimoto, Cosine (Orchini), and City Block similarity metrics. These techniques have been tested on a new Job Recommendation Systems Dataset taken from Kaggle. We have also analyzed the performance of similar techniques involving other distance measures, such as Euclidean distance. The performance of these similarity score-based techniques for generating the highest score-based recommendations is assessed using different evaluation metrics such as Accuracy, Precision, Recall, and F1-score respectively.