Verifying Classifications of Online Job Postings using Machine Learning

Abstract

Online job postings are a special type of text document. They form a mixture between advertising of the company and lists of requirements and responsibilities.

At Experteer job postings are a core part of their career network for candidates, recruiter and headhunter. To provide better results for all parties involved in a job posting each posting is assigned to a function, industry and career level. These classifications are important especially in context of a job search. That means a high quality regarding the classification is necessary.

With the help of machine learning methods it is possible to not only verify the correctness of these classifications but also deliver metrics about the quality of the data connected to the job posting. During this project Support Vector Machines where used to analyse the title and description of more than 10.000 job postings for each type of classification.

Feature selection and feature structure turned out to have the biggest impact on the verification quality while analyzing the functions of the job postings. Advanced parameter selection algorithms can improve the signal to noise ratio when using smaller sized feature vectors which in turn will lower the needed processing costs.

These trends can also be applied to verify other features of online job postings like the industry and career level classification used by Experteer.