Evaluation of Harezmi education model application data with text mining techniques

Lessons designed around a new education model called Harezmi Education Model (HEM) were implemented in 411 schools in Istanbul, involving 4195 students, for a period of 32 weeks during 2019-2020 academic year. Students were asked to reflect upon their experiences they had in these lessons on a form consisting of open-ended questions and the present study aims to examine and analyze the data in which students reflected their experiences of HEM practices. The hidden patterns acquired from data were revealed through Latent Dirichlet Allocation (LDA) method. The number of topics consisting of three corpuses determined around "What did you do? (Q1)", "What did you learn?" (Q2) and "How did you feel? (Q3)" questions were extracted with the Perplexity and Cv, a coherence score announced by Röder et al. (2015). The corpus of Q1 which has 63,938 documents was modeled for 54 subjects with 2,000 iterations and reached a coherence value of 0.73. The corpus of Q2 which has 69,482 documents was modeled for 47 subjects with 2,000 iterations and reached a coherence value of 0.79. The corpus of Q3 which has 36,465 documents was modelled for 54 subjects with 2,000 iterations and reached a coherence value of 0.70. Analyzing the length of the reflections of the students, it is revealed that the longest reflections were made by secondary school students and the second longest reflections were made by high school students. The shortest comments, on the other hand, were made by primary school students. Based on dictionary-based sentiment polarity analysis on Q3 corpus, primary school students are the happiest group followed by middle and high school students. In addition, analyzing Q1 topic labels, it is acknowledged that 5E Instructional model phases (engagement, exploration, explanation, elaboration and evaluation) as well as Computational Thinking Phase were labeled by experts. Furthermore, analyzing Q2 topic labels, it is ascertained that the main focus points of the lessons were environment, computational thinking and problem solving. Finally, analyzing the findings of Q3, it is discovered that students reflected upon their feelings as well as social norms. Within the scope of the current study, text mining and modeling methods were analyzed comprehensively in order to create a road map. By the revealed dataset which is the product of the study carried out and the patterns which are the reflection of students' perspective it can be conceivable that the study could be a basis for future studies in this field.