AI-BASED RESUME SKILLS EXTRACTOR AND RECOMMENDER MODULES FOR STATE UNIVERSITY HUMAN RESOURCE ANALYTICS SYSTEM
| Autor: | Paul Nigel S. Abalos, Thelma D. Palaoag |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2024 |
| Předmět: | |
| Zdroj: | Proceedings on Engineering Sciences, Vol 6, Iss 4, Pp 1473-1480 (2024) |
| Druh dokumentu: | article |
| ISSN: | 2620-2832 2683-4111 |
| DOI: | 10.24874/PES06.04.006 |
| Popis: | This quantitative-developmental research primordially seeks to develop two artificial intelligence (AI) agents, namely Skills Extractor and Job Recommender AI Agents, that will be involved in the skills extraction from a portable document format resumé (PDF) file and offer job recommendations via a larger web application-based data-driven state university human resource analytics system being concurrently developed also by the researchers. Dataset for this study which covers ___ resumes/applicants and ____ job posts in ___ job sectors came from the resumés and job posts from the largest state university in the Ilocandia Region of the Philippines. The researchers explored and compared Apriori data mining association rule algorithm and content-based filtering (CBF) approach to match extracted words and phrases from a resumé to skills banks and job posts and generate list of soft and hard skills using support, confidence, and lift metrics for the Apriori algorithm and cosine similarity score for the CBF algorithm, and from there generate job or applicant recommendations. For the validation, the researchers employed offline evaluation method by using relevancy approach through decision support metrics (accuracy, precision, recall, and F1-score), and ranking-based metrics (average precision or AP@k). Experimental results of the study have shown that the CBF algorithm has outperformed the Apriori algorithm which obtained mean accuracies of 92.30% for skills recommendation and 91.82% for job across ___ test job sectors. Meanwhile, significant mean accuracy differences of 4.28% in skills recommendations and 5.38% in job recommendations, respectively, were measured between the 2 algorithms in favor of the CBF algorithm. |
| Databáze: | Directory of Open Access Journals |
| Externí odkaz: |
|