JobMining@HfTL – an innovative project to identify training needs in ICT companies
The very short ICT technology cycles in today’s age pose a problem for many businesses: how to provide the right amount of training, with the right topics and at the right time. This problem is exacerbated by constant downsizing of budgets dedicated to further training. Educators are faced with the challenge of identifying important topics early on, to evaluate economic aspects, and to develop selected topics into educational products.
To tackle this issue, the JobMining@HfTL research project, which is led by Prof. Dr. Frank Bensberg, was inaugurated at the university. The main objective is to gain knowledge about education needs by analysing job advertisements posted on the Deutsche Telekom career site. By closely examining these open positions, it is possible to gather detailed information about which qualifications the applicants need to have. This information allows educators to plan and market their training programmes more efficiently and to focus on what is really needed.
The project team had to process mostly unstructured data using statistical and linguistic analysis methods. The digital source data had been presented in several different formats. Using a crawler, the team was able to extract a set of raw data and compile it into documents for further processing. Next, the contents were dissected into grammatical word types and filtered to create an index with classifiable keywords.
Within three months, over 1,100 job advertisements had been processed using the method described above. The raw data was used as the basis for text mining analysis. The advertisements were scanned for keywords relating to engineering or business study programmes (e.g. telecommunications and electronic engineering, computer science, business information systems, business administration). This step was supported by the text mining system, which operated using a query language allowing complex search expressions.
As a result of the comprehensive analyses, the project team was able to gain valuable insights into which ICT skills are currently in demand. This information is particularly useful for educational institutions where programmes are to be updated or completely revamped.
IT-related degree programmes in particular have to be balanced between conceptual knowledge, which is not subject to a lot of changes over time, and product knowledge, which can be very short-lived. In order to help find the right balance, the identified keywords were divided into three categories according to the type of training they represent, which are: concepts/methods, languages, and manufacturers/products.
• Concepts and methods include the basic working skills of potential applicants, for example in project management.
• The languages category includes programming skills, with Java currently playing a major role.
• The category manufacturers & products focuses on the use of office software such as Microsoft Office, individual applications, and SAP.
Out of all the new insights gained by text mining, those relating to languages and manufacturers/products were considered to be especially enlightening and of high potential value. It became apparent that in the ICT sector there are currently certain very important issues. One of the main ones concerns tools used to make the way information processing in businesses is organised more efficient, for instance the IT Infrastructure Library as a reference model for organising data centres.
Another problem which arose during the data analysis was that no predictions could be made regarding the development of individual topics. However, since it is known that ICT is a field where new concepts and technologies are adopted very quickly, the team was able to produce a trend analysis which showed the frequency of keywords over time and clearly marks topics with an unusual increase in frequency. This model can be used to visualise how the importance of specific skills/topics changes over time.
The analysis results have shown that text mining allows educators to extract interesting data from job listings and use it to modify the courses they offer. The HfTL is currently planning to introduce a part-time and a co-op Master degree programme in Business Information Systems. Information provided by the JobMining@HfTL project will be used in designing the course contents for these programmes and will therefore ensure that what is taught is in line with market demand.