Due to the spread of educational management information systems (EMIS), it become necessary to add intelligent layers to improve the educational process. One of the important tasks when the student moves from one stage to the other within the educational system of a university is the determination of the appropriate department if the transition is from the first level of a faculty to a certain department or the determination of the specialization track within a certain department in higher levels. These transition moments are crucial because they affect the degree of success of the student in the selected specialization and the quality of the educational process as a whole. In this research, different machine learning (ML) techniques have been tested to predict students' marks based on their marks in the preceded courses to guide them in the selection of the most suitable specialization or track. A variety of ML prediction models have been studied, experimented and evaluated on a propriety dataset, which resulted in the selection of a neural network (NN) architecture that gives an average root mean squared error of 6.26 and a mean absolute error of 5.74 based on a scale of 0 to 100. The accuracy is comparable to the state-of-the-art work and a practical example has been given that proves the ability of the proposed system to recommend certain tracks and/or specializations based on the marks of the already studied courses. Moreover, indirect prediction using cascaded networks has been proven to generate acceptable results that can facilitate building a hierarchy of networks using a short-term dataset to draw a weighted course road map that helps students to select the best path and help institutions to perform early measures to deal with weaknesses and anomalies.

Marks prediction neural network regression linear regression logistic regression support vector machines.