Designing a Bidirectional Job Matching Model Using Artificial Intelligence: the Case of Ministry of Labour and Skills (MoLSs)

EasyChair Preprint 15309

Abstract

The rapid advancement of AI and big data presents both challenges and opportunities for global labor markets. Urban youth are highly plagued by frictional unemployment. This incredibly high level of frictional unemployment has spatial and informational components, driven by high costs. However, currently existing but undiscovered matches between job seekers and open positions account a smaller portion of urban unemployment. This study explores leveraging digital tools to enhance the effectiveness of Ethiopia's public employment services, which face limitations in job matching, labor market information, and policy implementation. The research aimed to design an AI based job matching model to improve the public employment service support provided by Ethiopia's Ministry of Labor and Skills. The study followed a hybrid knowledge discovery process (KDP) data mining model, which includes understanding the problem domain, collecting historical data, preparing datasets, and four logically designed experimentations for selecting best fit job matching model for the public employment service. The designed job matching model match and suggest the best fit jobs to the jobseeker or vis versa by calculating a similarity index using vector representations job seeker's and the available job postings features in the dataset to describe features (documents and corpus of texts) using n-dimensional vectors which each dimension representing the frequency of a certain term in a document in the available skill sets and then ranking them according to their Doc2vec algorithm and cosine similarity measure. This designed model has been implemented in python. The selected Doc2Vec model with selected text embedding’s from job seeker and vacancy datasets achieved a cosine similarity of 0.99 to 1, revealing the desired result for the best-fit job matching model. The high cosine similarity result suggests the model is performing well in matching jobs to the jobseeker.

Keyphrases: AI, Frictional unemployment, Hybrid KDP

@booklet{EasyChair:15309,
  author    = {Tewodros Desta and Million Meshesha},
  title     = {Designing a Bidirectional Job Matching Model Using Artificial Intelligence: the Case of Ministry of Labour and Skills (MoLSs)},
  howpublished = {EasyChair Preprint 15309},
  year      = {EasyChair, 2024}}