Efficiency of Decision Tree Depth to Diagnose Mathematical Procedures in Number and Algebra for Seventh-grade Students

This study is dedicated to design and compare the efficiency of predictive models with different depths of decision trees for diagnosing mathematical procedures in number and algebra. Leveraging a dataset comprising responses from 509 seventh-grade students across 4 regions in Thailand, the investigation applies the Python programming language with decision tree methodologies to construct predictive models capable of accurately determining students' proficiency in mathematical procedures. The findings of the research elucidate that model efficiency is contingent upon the depth of the decision tree. Notably, the model with a tree depth of 5 surpasses others in efficiency, evidencing an accuracy rate of 0.87 and an error rate of 0.37. Meanwhile, the model with a tree depth of 4 demonstrates slightly lower efficacy, with an accuracy rate of 0.76 and an error rate of 0.51. These results underscore the critical influence of decision tree depth on the performance of predictive models within the educational domain.