Abstract:

A university curriculum is generally flexible. The study program to obtain a university degree is conformed by several courses, which are distributed in academic terms. Prior to the beginning of each academic year or term (depending on the organizational and academic rules), the student should enroll on one or more courses of which some are compulsory and other optional, corresponding to the period according to his/her progress, personal interests and preferences. On the other side, after making decisions about courses to be attended, students’ achievements and academic performance also depend on different kinds of criteria, ranging from personal nature for each student (e.g. preferring natural sciences, courses with more practical work etc), to other preferences presenting actions/attitudes/ behaviors of other students with approximately the same characteristics, such as academic performance, demographic characteristics or other personal information.

Accordingly, the consideration of more sophisticated preferences is essential. Human preferences are not simple, decisions about their desirable requirements can be dependent on other internal or external factors, and their modeling is a great challenge, as it is difficult to express human opinion in a way that can be easily processed by computers. In literature, many analyzes were taken in order to determine data available from learning environments and learning systems that can detect students' learning behaviors in real-world scenarios, examining students' motivation which influence students interest, persistence and performance in academics. Most of them showed different students' interests in online and traditional courses, time constraints due to work and home responsibilities are of high importance in decision making, interaction of demographic characteristics among motivation orientation, etc.

This paper proposes innovative approach for making predictions of the most preferable courses based on student's preferences and simultaneously making estimations on students’ academic performance by considering previous students’ achievements and their shown attitudes. The main contribution of proposed approach is a framework for representing and processing of different kinds of preferences of each individual student based on the well-known and widely adopted Analytic Hierarchy Process (AHP). This adoption is used for defining measurements needed for making quantitative models over courses process workflow and its optimal configuration.

Proposed model is developed based on promising results of using quantitative metrics for optimal configuration of families of business processes recently developed and explored in Software Product Line Engineering discipline. This innovative approach of transferring developed methods and techniques to learning environment is a result of the national project titled ‘Development of dynamically adaptive service-oriented architectures (SOA) based on non-functional requirements and historical data’ funded by Ministry of sciences, Montenegro. The project was aimed on developing methods and techniques that will enable development of SOAs based on various types of non-functional requirements and additional requirements obtained by analyzing available historical data, with further practical applications in both, learning and medical environments.

Keywords:

Predictive model, students' preferences, academic performance, AHP.