Association Rule Mining for Uncovering Co-occurring Skill Sets in Multi-Portal ICT Job Advertisements

Abstract

Rapid changes in the ICT labour market require stakeholders to understand not only which skills are demanded but also how these skills occur together. However, existing studies provide limited insights into co-occurring skill demands in ICT jobs. This study addresses this gap by uncovering skill sets that frequently appear together in ICT job descriptions. A total of 15,653 unique ICT job advertisements published between 2020 and 2025 were collected from LinkedIn, AjiraYako and Mabumbe to capture regional and international contexts. Guided by the Cross-Industry Standard Process for Data Mining (CRISP-DM), association rule mining using the Apriori algorithm identified frequent itemsets and generated rules evaluated using support, confidence and lift. The analysis revealed strong co-occurrences among programming skills, such as C# and C++ (confidence = 1.00; lift = 2.41), although this combination occurred in a small proportion of advertisements (support = 0.0096). Java and Python showed the highest support (0.1818), with confidence values of 0.51 and 0.42 and a lift of 1.19, which indicate a positive association despite belonging to different application domains. By uncovering skill associations beyond individual skill frequencies, the study provides richer labour market intelligence for workforce development and policy formulation.