Description

To make a training program for unemployed people, the government wants to understand what hard skills a person can acquire, to get a job, based on the experience one already has. 

The project identifies the key hard skills for different job categories, builds models that could predict the most suitable job category and compares the performance. Job descriptions are used as a proxy for the candidate experience due to the data privacy issue.

Job ads text was used as a proxy for the candidate experience. Fortune 500 job ads were analysed.TF-IDF was applied to extract the keywords for hard skills, which could be predictors of the job title for each O*NET category. Logistic regression, Random Forest, and Neural Networks were considered. Logistic regression showed the best Accuracy score and best weighted F1. Logistic regression is also the most transparent one, making it easier to identify which hard skills contributed to the job category receiving high probability score.