Abstract:

Master students of the Graz University of Technology form their master curriculum by selecting courses from a number of catalogues. Selection from one catalogue is restricted, and the only criteria are the total number of ECST that give selected courses. The curriculum assistant is a special component of the campus management system that provides students with automatically generated variants of the curriculum.

Deep learning is the latest technology that considerably extends the opportunities of machine calculation. Deep learning is successfully used in many areas such as image recognition and others. Deep learning is expected to allow machines to solve problems in a manner similar to the human way of thinking.

In this paper we describe an innovative application applying deep learning technologies to facilitate the curriculum assistant with a number of artificial intelligent features. Specifically, we introduce to the curriculum assistant an evaluation of perspectives to get a particular type of job after the university. Thus, the intelligent assistant infers the particular curriculum not only taking into account total number of ECST and necessary courses diversity but also skills provided by the courses and current demand for such skill on the recruitment market.
This project is intended to develop an application that gathers the recruitment data from announcements, analyze the data using principles of machine learning, and infer study curriculums using different criteria.

Thus, the application is able to infer
• the most demanded study curriculum,
• the best study curriculum for a particular job,
• the most well-paid study curriculum, etc.

Additionally, the system provides some visual analytics like distribution of lob offers over some geographical area.
The project is being implemented on the base of Azure Machine Learning services and consists of 3 sub-systems.
Internet crawler: the system scans all known websites with lob announcements, parses the documents and returns a unified JSON file representing parameters of each announcement. The most challenging part of this component that requires deep learning methods is identifying all the required skills and presenting them in a unified form.

Intelligent Backend: the component that:
• gathers results of the crawler and stores data on all the job vacancies into the special database with possible modification of the existing info.
• teach the special neural net using the skill parameters and the salary offers.

Graphical Front-End: the component implements a GUI that provides end-user access to the system.
Thus, the system may infer the study curriculum providing the skills that are most demanded. The system can predict a salary for a particular combination of skills. The system may infer a curriculum that is best for getting job in a certain geographical area, or that assure highest salary expectations.

From the university authority perspective, the system may automatically identify new knowledge and skills that are in demand on the job market but not provided by university courses.

Keywords:

Deep Learning, Knowledge Management, generating Curriculum.