Automated model to rank candidates for a job position based on data extracted from LinkedIn profiles

Abstract

Recruitment process has become very crucial in the vast job market and being able to recruit effectively is a challenge. This is because, there is a scarcity of suitable candidates for any particular job opening. Moreover, the ratio of a suitable candidate to the number of job opening is very low. Therefore, many multinational companies are investing fortunes in their recruitment teams. The only information that the recruiters have during the process of recruitment is the curricular vita, based on which a short interview is scheduled and then the candidates are hired. This does not give the recruiters an insight to their skills and educational background in a single format as different people write CVs in different ways. Also, recently, in most curricular vita the social handles such as Facebook or LinkedIn are provided. Data in these platforms can be taken advantage of to find information which are essential to ensure an efficient and successful recruitment. These data collected can be analyzed to match with the job requirements resulting in a more accurate recruitment process with data driven decision making. The two major entities in this process are the recruiters and the candidates who applied for the job. The challenge is to find a qualified candidate for a particular job that fulfills all the requirements of the job. Therefore, in this paper we have collected a data set of approximately 300 candidates, automatically, from their LinkedIn profiles for a job of a Software Engineer. Then, we have used NER of BERT model to pre-train the dataset – to summarize the text using NLP. Then, we have used the VADER model to carry out sentimental analysis of the text data. After that, we weighted each entities namely: About, Skills, Education Background, Experience and Language. Priority of each attributes were carefully considered by experts at Bangalink Digital Ltd. according to which they are weighted. Then, using XGBoost Machine Learning Algorithm, we have trained the system. Finally, we have used the TOPSIS Algorithm to rank the candidates and have a holistic idea of the quality of the applicants in a descending order of priority.