Abstract:

Occupational titles such as International Standard Classification of Occupations (ISCO) are severely unsuitable to measure skills since they might take different meanings and interpretations. In a scenario where on-the-job training takes priority over formal schooling (i.e. creation of new workplaces due to rapid and unexpected technological changes), using educational proxy for skill analysis is equally questionable. In more general terms, a pivotal issue faced by employers, jobseekers and policy makers dealing with planning education and training strategies is that there are significant differences in skills demand for relatively similar jobs with different job titles, which, as a result, often fail to reflect the true nature of the work. The need to flexibly derive competency requirements in a way that bypasses rigid occupational schemes might be addressed by utilizing machine learning and digital vacancy data. The focus of our research is venture capital and private equity jobs, with opportunities for roles ranging from analysts and associates to directors. Private equity and venture capital jobs and careers fall into several categories, including analysts at the junior end, who build financial models and crunch numbers to help establish whether an investment is viable; principals who play a key role in appraising whether to do a deal and managing the process if it goes ahead, and partners who lead the fund, seek new investments and liaise with investors. Funds also employ investor relations and operations professionals. Online recruitment portals offer huge volumes of vacancy data and provides unexplored opportunity for analysis. The efinancialcareers job portal lists more than 90 different job titles for private equity/venture capital vacant positions.

Our research is aimed at outlining a certain procedure that refers to the existing occupational settings and might successfully support in mining thorough and actionable information. We developed and tested a method to pursue two distinct objectives. First, automate the process of sorting vacancies into occupational groups based on job titles by separating vacant positions based on the analysis of job descriptions that offer richer and more detailed information, although in a more unstructured fashion. Secondly, in order to better describe and profile skills in demand, the terms’ occurrences are gathered in word vectors and visualized in intuitive and straightforward fashion.

The procedure we followed starts with dataset exploration, which revealed an unsurprising variety in length, structure, and type of information, and data cleaning using replacement techniques on the “Job Title” and “Job Description” wordings. The next step is data preparation whose aim is to create a labeled dataset of examples. Data modelling for classification purposes (using K-Nearest Neighbors) is carried out to assign the labels quickly and efficiently in the cases where job titles are ambiguous. The final step is data visualization via word clouds that are a reasonable representation of job descriptions. Each distinct category includes a set of skills both of a technical nature (i.e. direct investment research) and of a soft one (i.e. team working, broad communication). The potential of the procedure is discussed in terms of translating the outcome into training provision, and guiding learners with soft and hard competencies accordingly to their chosen careers.

Keywords:

Skill needs, venture capital, private equity.