Globalisation and technological changes are bringing about new trends in the labour market and challenges for policy-makers, businesses and workers. Anticipating skill needs and being prepared to meet them effectively is one of the leitmotifs of the political agenda.
The 2013 labour authority VET reform specifies that public training provision has to supply the skills that meet productivity and competitiveness requirements of enterprises and challenges arising from changes in production systems. It also needs to support career and personal development and worker employability.
To plan and design public training programmes, in the short, medium or long term, it is important to have accurate information on skill needs and how skills are linked to different jobs. Many different methodologies are being explored. The State Foundation for Training in Employment (Fundación Estatal para la Formación en el Empleo) is testing big data techniques, taking as starting point the learning contents requested by companies for publicly subsidised courses, assuming these are good proxies for companies’ needs.
The classification system used so far groups together training activities with similar content – which can be presented for funding under different names or delivering modes – associating them to the national classification of occupations (the CNO-11, within the conceptual framework of ISCO-08).
In 2015, for instance, on-demand VET provision (training programmes initiated at the company’s request and subsidised by the state) supported 415 000 training actions. These included a wide range of skills which may have had different names but are similar in learning content.
Big data technologies, such as data mining, clustering and text analysis, use machine learning methods and algorithms to identify patterns in learning content for the requested training action. They are yielding promising results in terms of efficiency and quality of the training offer.
This system can generate classification rules using decision trees that can be applied to training activities, assigning to each of them a probability of belonging to a category. The system is scalable over time, allowing the classification of future actions and identifying new trends.
Big data techniques, if combined with qualitative methods and other data sources can substantially support skills anticipation.