Forecasting of successful completion of university study programs: data pre-processing and optimization of LAMA BPO algorithm

Abstract

Lithuanian school graduates wishing to be admitted to state-funded places at universities undergo a competitive selection based on their final school and state exam grades. The problem of organizing competitive selection is that in Lithuania there are different types and scales of school knowledge assessments. Algorithm developed by LAMA BPO address this problem by adjusting grades into a single scale. But choice of final arithmetic values into which pupil’s grades are converted is not justified theoretically. Proposed by the authors algorithm is a development of the LAMA BPO algorithm and allows to achieve a consistently higher accuracy of predicting learning results at the university. The higher accuracy of the models indicates a better capture of the central trend: a positive correlation between the level of performance in individual school disciplines and the results of university education in certain study programs.