Search

Curriculum mining is a recent research area that applies a data-driven approach to assess students' learning behavior by discovering the curriculum model and compare it with the curriculum guideline. Some previous works exist to cope with the problem to discover the curriculum model from student database, by utilizing the concept of process mining. However, the challenges of discovering the curriculum model remain due to the different nature of student database from event log in twofold; the level of time granularity and variability of instance attributes. Previous works on curriculum mining that deal with conformance checking of the model between student's learning behavior and curriculum guideline are related to sequence matching alignment which is insufficient to understand the patterns of a group of students in a particular level of time granularity, i.e., Semester. This study proposes a curriculum mining methodology for curriculum assessment and students' learning behavior by creating segmentedtrace profiles. The segmented-trace profiles are extracted based on a local alignment of sequences and generated as the input for sequence matching alignment to assess whether the observed students' learning behavior match with the prior curriculum guideline. The profiles would be the features of the clustering approach. Real curriculum data has been used to test the effectivity of the methodology. The results show that the students can be grouped into various clusters per semester that have different characteristics for their learning behavior and performance. The results can be analyzed further to improve the curriculum guideline. [ABSTRACT FROM AUTHOR]