skip to main content

Title: Detecting Trait versus Performance Student Behavioral Patterns Using Discriminative Non-Negative Matrix Factorization
Recent studies have shown that students follow stable behavioral patterns while learning in online educational systems. These behavioral patterns can further be used to group the students into different clusters. However, as these clusters include both high- and low-performance students, the relation between the behavioral patterns and student performance is yet to be clarified. In this work, we study the relationship between students’ learning behaviors and their performance, in a self-organized online learning system that allows them to freely practice with various problems and worked examples. We represent each student’s behavior as a vector of highsupport sequential micro-patterns. Then, we discover both the prevalent behavioral patterns in each group and the shared patterns across groups using discriminative non-negative matrix factorization. Our experiments show that we can successfully detect such common and specific patterns in students’ behavior that can be further interpreted into student learning behavior trait patterns and performance patterns.
; ;
Award ID(s):
Publication Date:
Journal Name:
The Thirty-Third International Flairs Conference
Sponsoring Org:
National Science Foundation
More Like this
  1. Educational data mining research has demonstrated that the large volume of learning data collected by modern e-learning systems could be used to recognize student behavior patterns and group students into cohorts with similar behavior. However, few attempts have been done to connect and compare behavioral patterns with known dimensions of individual differences. To what extent learner behavior is defined by known individual differences? Which of them could be a better predictor of learner engagement and performance? Could we use behavior patterns to build a data-driven model of individual differences that could be more useful for predicting critical outcomes of the learning process than traditional models? Our paper attempts to answer these questions using a large volume of learner data collected in an online practice system. We apply a sequential pattern mining approach to build individual models of learner practice behavior and reveal latent student subgroups that exhibit considerably different practice behavior. Using these models we explored the connections between learner behavior and both, the incoming and outgoing parameters of the learning process. Among incoming parameters we examined traditionally collected individual differences such as self-esteem, gender, and knowledge monitoring skills. We also attempted to bridge the gap between cluster-based behavior patternmore »models and traditional scale-based models of individual differences by quantifying learner behavior on a latent data-driven scale. Our research shows that this data-driven model of individual differences performs significantly better than traditional models of individual differences in predicting important parameters of the learning process, such as performance and engagement.« less
  2. Hawkes processes have been shown to be efficient in modeling bursty sequences in a variety of applications, such as finance and social network activity analysis. Traditionally, these models parameterize each process independently and assume that the history of each point process can be fully observed. Such models could however be inefficient or even prohibited in certain real-world applications, such as in the field of education, where such assumptions are violated. Motivated by the problem of detecting and predicting student procrastination in students Massive Open Online Courses (MOOCs) with missing and partially observed data, in this work, we propose a novel personalized Hawkes process model (RCHawkes-Gamma) that discovers meaningful student behavior clusters by jointly learning all partially observed processes simultaneously, without relying on auxiliary features. Our experiments on both synthetic and real-world education datasets show that RCHawkes-Gamma can effectively recover student clusters and their temporal procrastination dynamics, resulting in better predictive performance of future student activities. Our further analyses of the learned parameters and their association with student delays show that the discovered student clusters unveil meaningful representations of various procrastination behaviors in students.
  3. In recent years, research has associated grade point average (GPA) with a variety of student outcomes during their undergraduate careers. The studies link higher GPAs to students being more likely to graduate in their major, while lower GPAs have been linked to students switching majors or leaving the institution. Further research, which focuses on how Black female and male students remain successful in different engineering degrees, is necessary to identify the underlying elements contributing to their entrance into and exit from engineering disciplines. This quantitative examination of trends among the GPAs of Black women and men is part of a larger NSF-funded mixed-methods study that includes in-depth student interviews of Black students who persisted in and switched from ME. In this quantitative paper, we examine the GPA patterns of Black students in Mechanical Engineering (ME). Students who have ever enrolled in ME have four potential, mutually exclusive, outcomes: 1) they can persist for 12 semesters without graduating; 2) they can graduate in ME within 12 semesters; 3) they can switch to another major; or 4) they can leave school. In this research, we identify the most common GPA patterns associated with graduated ME students. We hypothesize a relationship between distinctmore »GPA patterns and whether a student persists in ME, graduates in ME, switches away from ME, or leaves the institution altogether. This quantitative investigation uses the Multiple-Institution Database for Investigating Engineering Longitudinal Development (MIDFIELD) to collect the cumulative GPA of ME students at each term. We use a functional cluster analysis approach to group similar patterns. First, a function is fit to each student record. Then a cluster analysis is conducted on the function parameters to identify natural groupings in the data. Once students are grouped according to their GPA profile, we examine the other characteristics and outcomes of the group. We present a visual quantitative analysis of the patterns in the GPAs of Black women and men who enroll in ME. Clustering analysis suggests that first-time-in-college (FTIC) Black female students in ME who graduated have a higher proportion of students in the higher GPA clusters than the proportion of FTIC Black male students who graduated in ME. A higher proportion of the male student population is clustered in the lower GPA cluster groups as compared to women in the lower GPA cluster groups. A higher proportion of students who graduated are in the higher GPA clusters than the proportion of graduated students in the lower GPA clusters.« less
  4. As computer-focused policies and trends become more popular in schools, more students access math curriculum online. While computer-based programs may be responsive to some student input, their algorithmic basis can make it more difficult for them to be prepared for divergent student thinking, especially in comparison to a teacher. Consider programs that assess student work by judging how well it matches pre-set answers. Unless designed and enacted in classrooms with care, computer-based curriculum materials might encourage students to think about mathematics in pre-determined ways. How do students approach the process of mathematics while using online materials, especially in terms of engaging in original thought? Drawing on Pickering’s (1995) dance of agency and Sinclair’s (2001) conception of students as path-finders or track-takers, I define two modes of mathematical behavior: trail-taking and bushwhacking. While trail-taking, students follow an established approach, often relying on Pickering’s (1995) disciplinary agency, wherein the mathematics “leads [them] through a series of manipulations” (p. 115). The series of manipulations can be seen as a trail that a student may choose to follow. Bushwhacking, on the other hand, refers to actions a student takes of their own invention. It is possible that, unknown to the student, these actions havemore »been taken before by others. In bushwhacking, the student possesses agency, which Pickering (1995) describes as active (rather than passive) and as hallmarked by “choice and discretion” (p. 117). In this study, students worked in several dynamic geometric environments (DGEs) during a geometry lesson about the midline theorem. The lesson was originally recorded as part of a larger study designing mathematically captivating lessons. Students accessed both problems and online addresses for corresponding DGEs via a printed packet. Students interacted with the DGEs on individual laptops, but were seated in groups of three or four. Passages of group conversations in which students transitioned between trail-taking and bushwhacking were selected for closer analysis, which involved identifying evidence of each mode and highlighting the curricular or social forces that may have contributed to shifts between modes. Of particular interest were episodes in which students asked one another to share results, which led to students reconsidering previously set approaches, and episodes in which students interacted with DGEs containing a relatively high proportion of drag-able components, which corresponded to some students working in bushwhacking mode, spontaneously suggesting and revising approaches for manipulating the DGE (e.g., “unless you make this parallel to the bottom, but I don’t think you... yes you can.”). Both types of episodes were found in multiple groups’ conversations. Further analysis of student interactions with tasks, especially with varying levels of student control and sharing, could serve to inform future computer-based task design aimed to encourage students to productively engage in bushwhacking while problem-solving.« less
  5. Modern online learning platforms offer a wealth of learning content while leaving the choice of content for study and practice to the learner. Recent work has demonstrated that many students use inefficient learning strategies that lead to lower performance in this context. The ability to detect inefficient learning behavior by monitoring learning data opens a way to timely intervention that could lead to better learning and performance. In this work, we propose SB-DNMF, a structure-based discriminative non-negative matrix factorization model aimed to distinguish between common and distinct learning behavior patterns of low- and high-learning gain students. Our model can discover latent groups of students' behavioral micro-patterns while accounting for the structural similarities between these micro-patterns based upon a weighted edit-distance measure. Our experiments demonstrate that SB-DNMF can find meaningful latent factors that are associated with students' learning gain and can cluster the behavioral patterns into common (trait), and performance-related groups.