USE OF LINKEDIN ENDORSEMENTS IN RECOMMENDER SYSTEMS

Abstract

"Social networks have become an important source of information, especially professional networks, where users share information about their academic and professional qualifications, skills and work experience. Nowadays, where the updating and development of professional skills is becoming more and more relevant for professionals, this information is of great interest, since it allows to know the trend of the labor market. In this regard, LinkedIn, in particular, has become one of the most widely used professional networks for this purpose, designed for professional networking and job search. From the professional profiles shared in this media, it is possible to retrieve relevant information for the labor sector, to know information about the professional profiles according to their competencies, as well as the most demanded competencies in the different job positions. This makes it possible to detect formation needs to improve or develop new skills. Additionally, LinkedIn has a particular element, the endorsements, through which it allows members of the network to acknowledge the skills of other members, which could provide information related to the level of development of a given skill. The analysis of this information, in addition to detecting training needs, can be used to adapt curricula to meet these needs, as well as in the field of human resources, to find the right candidates for the job. Currently, recommender systems have become a powerful tool for suggesting relevant articles to users. In the field of education, they have become very powerful, making it possible to link the training offer with the training needs of users, especially in the field of continuing education, in order to meet the need to develop professional skills. In a previous work, we have developed a recommendation system based on machine learning and ontology to recommend continuing education courses to LinkedIn users. As an extension of our work, we propose to incorporate the endorsement information to the user profiles to determine the improvement in the recommendations of our recommendation system. The results obtained showed an improvement in the recommendations, obtaining an accuracy of 94%."