In the master's thesis we present an analysis of the results in the Bebras competition. The main purpose of the master's thesis is to identify factors that significantly affect the prediction of the final results of the competitor in the Bebras competition. For this purpose, we use machine learning methods.
In the theoretical part of master's thesis we described the concepts of computational thinking. The definition is presented in accordance with the developmental definitions that best described its meaning and applicability of the term in its particular periods of development. Our research is based on operational definition of computational thinking for K-12 educators, which highlights the charateristics of the problem-solving process. We describe the computational concepts underlying the definition of computational thinking. Understanding them is important in the analysis of collected data. In the process of predictive modelling we used the concepts of computational thinking and domains or fields of computer science as a predictive variables. The definition of computational thinking is followed by a description of computer science as a part of the general school curriculum. We described the current situation in Slovenian educational system regarding to teaching elementary and secondary computer science courses or content. In recent years significant changes have been made in the field of computer science education, where the K-12 curriculum for computer education is at the forefront. The K-12 curriculum defines the conceptual guidelines and frameworks for computer science education from kindergarten through high school. In accordance with the suggestions of the K-12 curriculum, in the thesis we present computational practices and computational concepts that students should be familiar with during their education.
An important area of computer science is the development of computational thinking. In the thesis, we describe it through the prism of Bebras competition, which is intended to promote computational thinking among elementary and secondary school students. In line with the development of the importance of computational thinking and the increasing popularity of the Bebras competition, different classifications of the concepts that define the term was created by different authors. Based on the different classifications, for the purposes of our research, we defined our own classification, which was used in the further data analysis.
In the empirical part of the thesis, we analyse the factors that most significantly influence the final achievement of the competitors in the Bebras competition. Using machine learning technique random forests, we designed predictive models to predict target varible based on various predictive variables. The design of the predictive models was based on an acquired sample of Bebras contestants in the 2018/19. We designed several predictive models for each competition group. As predictive variables we used the concepts of computational thinking, the computer science fields and the students' class. Predictive model of 6th and 7th grade competitors show that the most important predictive variables were the concepts of modelling and simulation and algorithmic thinking. Model built for 8th and 9th grades competitors, where the most important predictive variables were the concepts of modelling and simulation and algorithmic thinking. The model for 1st and 2nd grade secondary school competitors showed that the most important predictive variables were the class (age) that competitors attend and the concept of modelling and simulation. When considering all of the predictive models that were built, the most important predictive variables were the concepts of modeling and simulation and algorithmic thinking. Such an analysis provides a more accurate insight into the factors that have the most important influence in predicting the final result of a competitor in a Bebras competition.