| Abstract [eng] |
Each driving school has its own or uses test solution systems for road traffic regulations created by other producers which helps to learn driving theory in schools or solve tests using access codes from distance The main problem of these systems is repeated questions, when tests are solved several weeks. When the questions are repeated learners memorize the answers. Thus knowledge of learners gets worse and solution of the test do not show the real level of knowledge. The market analysis of test solutions systems of traffic regulation showed, that questions repeat when tests are solved repeatedly. Some questions were presented more often than others. It was analysed which type of the tests can be used to optimize presentation of test questions the automated test collection type was chosen, because various, vast collections can be created. For exam tests- computer adaptive test type was chosen, because of its progressiveness and real learners skills can be revealed. It was found out by comparing the algorithms, which can be used for generating test, that most of random number generating algorithms have systematical error. Due to this error some questions are presented more often than others. It was found out, that Fisher-Yates algorithm do not have systematical error. Four algorithms were implemented. Speed tests were performed and results were compared. It was proved, that Fisher-Yates algorithm does not have systematical error, because all questions are presented in the same frequency. Adaptive test type was applied for final skill tests. |
