Association between Test Item's Length, Difficulty, and Students' Perceptions: Machine Learning in Schools' Term Examinations

The study applies machine learning (ML) algorithms to investigate the association between length of a test item written in Chinese (through word count), difficulty, and students’ perceptions (IPs) science term examinations. For Research Question 1, items for grade 7 students aged 12–13 Taiwanese secondary school from 2014 2019 were analyzed. 2, included 4,916 said population. RQ3, gathered 48 same 2020. study’s results showed that first, average count 611 was 88.81, with an stem 41.16, options 47.66, stem-to-options ratio (S-O ratio) 1.27. Second, given ML M5P categorization algorithm affirms items’ predictive power, is key factor determining its difficulty. As result this algorithm, 3 categories examination classified (<57.5 words, 57.5–91.5 >91.5 words), linear prediction models difficulty (LM1, LM2, LM3) generated. From these models, it found as increases, so does In analysis IPs, J48 better convertible into understandable rules. IP root node decision rule, indicating importance variable. Therefore, would more likely answer correctly when 1) perceived be easy or normal, 2) had high ordinary achievements science, 3) contained less than 71 words. can used reference educators, examiners, researchers practical design. Moreover, guide further research method direction applying analyzing scientific assessments.