Abstract
This study aims to develop a data mining framework to analyze key factors of unemployment duration and the complex relationships among each factor. Conducted on the basis of real data collected from a representative Human Resource Agency in Taiwan. In order to extract latent knowledge and patterns from huge data about job seekers. This study formulates research hypotheses based on literature review, domain expert knowledge and supported by Bayesian Network, statistical test and correlation coefficient to screen out 15 key factors of the job seekers’ “general information, job requirement, education, working experience and new work category” have a significant effect on unemployment duration. Then, using Bayesian network to clarify the relationships among each factor and unemployment duration. Finally, this study presents a process of case studies that can extract the useful knowledge of data mining results efficiently. Major findings indicate that unemployment duration difference among each field, the employment tendency of different type of job seekers and job transition patterns in the current domestic labor market. For example, in a particular industry, workers with 3 to 6 years’ seniority may have a high turnover intention, the reemployment difficulty among middle aged workers, the regional wage gaps and other social issues. The results assist various types of job seekers obtain comprehensive information to find their own niche in the labor market. In the meantime, this study also provides the decision-making reference for government and enterprise. On the other hand, Human Resource Agency can base on the results to improve their services. Help job seekers to find the most favorable direction.