A Job Recommendation System Based on Skill Importance and Deep Learning Label Prediction

The rapid growth of online recruitment and job searching platforms has enabled businesses and applicants to review large volumes of offers with a click of a button, allowing them to select the most suitable offers based on limited information. Job recommender systems, combined with skill extraction algorithms applied to both job descriptions and resumes, can automatically suggest the most suitable jobs to applicants. However, limitations-such as recruiters or applicants omitting certain skills, either due to lack of knowledge or considering some information to be trivial-can lead to incomplete representations of both the job and the applicant's abilities. Additionally, variations in resume structure may cause crucial information to be overlooked, negatively impacting the effectiveness of recommender systems. In this paper, we employ a state-of-the-art tool to extract skills from resumes, regardless of their structure. Then, we applied Long Short-Term Memory (LSTM) deep learning to predict resume labels, improving the preparation of flawed resumes for the recommendation phase. We propose an importance-based job recommendation system that recommends the most suitable jobs based on skill importance, which is calculated by the PageRank algorithm. This system matches applicants' skills with the skills required for each job using a weighted Jaccard similarity. Finally, our approach was evaluated on a dataset of 1,000 randomly selected computer science resumes, and the performance of the job recommendation system was measured using the nDCG@10 metric, achieving a score of 87.07% for job recommendations.