Behavior Analysis of Online Assignment Submission Patterns
Abstract
The primary aim of this scholarly investigation is to comprehensively
examine the correlation between students’ behavior in submitting assignments
via online platforms. Specifically, the study endeavors to scrutinize
three fundamental components: Procrastination, Penalties, and Cheating.
The principal inquiries guiding this investigation encompass the underlying
reasons behind student procrastination, the influence of adverse
academic performance and penalties imposed by instructors on students,
and the methodologies for identifying instances of academic dishonesty.
To achieve this objective, an Efficiency Index was developed, incorporating
methodologies such as Association Rule Mining, Kernel Density
Estimation, and basic Standard Deviation metrics. The findings reveal a
significant prevalence of student procrastination, particularly towards the
culmination of the academic year. Furthermore, the study suggests that
students exhibit a propensity to learn from their errors and experience
substantial improvement following negative feedback. Moreover, the
implemented algorithm successfully detects potential instances of academic
dishonesty by analyzing the Standard Variation of the student Efficiency
Index. Notably, high-achieving students display a lower likelihood of
transitioning to average performance compared to their lower-performing
counterparts, indicating a positive outcome.
6 Cheating
6.1 Introduction
According to Anderman and Murdock [5], the prevalence of academic dishonesty
within educational institutions is pervasive, emanating from diverse factors,
including an inadequate grasp of effective learning strategies or a reluctance to
invest requisite time in their implementation. Furthermore, the authors contend
that ”certain students resort to academic dishonesty due to their heightened
fixation on extrinsic motivations, particularly grades” ([5], p. 23).
One question that would be important to ask and answer before we move
forward is: Would it be reasonable to cluster the students based on the time it
took for them to finish the homework? Is there a positive correlation between
the amount of time taken for them to finish and their results?
It would make sense to believe that some students that have finished extremely
fast (for example, finished in 10 minutes when most finish in 2 hours)
might have cheated if they clearly got good grades? Is this a behavior that we
could spot based on how long this student usually takes in relation to others?
6.2 Data Manipulation and Testing
To answer this question, I have looked at an aggregated data that contains the
time taken for the students to finish their homework and their results. Then, I
have created a way of measuring their efficiency on doing homework by creating
the ”Efficiency Index”, which consist of a simple regression between the average
time it takes for a student to finish their homework and the average number of
minutes that it would take for them to improve their grade by 1 point.
As you can see in the Figure 5, The best way of aggregating this information
for better visualization is by using Kernel Density Estimation. KDE is a
non-parametric way to estimate the probability density function of a random
variable based on the observed data.
Each small red point is a student from our sample. The X axis is the time
taken to finish the assignment and the y axis is the grade they got in this
assignment. On the right and upper graphs , I show the density of each
characteristic.
6.2.1 Efficiency Index
I will analyze specific data pertaining to the Efficiency Index (EI). Initially, I
eliminate students who did not complete the academic year (dropouts) from
the dataset. Subsequently, I compute the EI for each student throughout the
semester. Following this, I determine the standard deviation of aggregate EI
values across students for each homework assignment. For instance, during
Homework 5 in 2020, the observed standard deviation was as indicated in figure 6.
6.3 Results
Noteworthy trends emerge from this analysis. The figure 6 graphically depicts
the temporal evolution of this index. It commences at a moderate level,
undergoes a decline, and stabilizes until the semester’s conclusion. Abruptly,
toward the semester’s end, a considerable spike in variability occurs.
This observation delineates an unforeseen disparity in the overall efficacy of
the class, stemming from a sudden and pronounced escalation in the productivity
of specific students. This phenomenon intimates a plausible upswing in
academic impropriety, notably manifested in a surge of efficiency toward the
semester’s denouement.
In examining the trends within academic data, the investigation aimed to
refute mere happenstance. To fortify the integrity of the analysis across multiple
academic years, the decision was made to eliminate outliers from both the 2021
and 2022 datasets. This rigorous curation involved systematically removing data
points extending beyond 3 standard deviations from the mean of the Efficiency
Index (EI) [17]. The resulting findings presented a consistent narrative woven
across each academic year, with Figures 7 and 8 showcasing a distinct rise in
standard deviation as each academic term concluded.
6.3.1 Other Possible Interpretations
However, delving deeper into academic integrity and the multifaceted nature
of student behavior, it’s crucial to consider various elements beyond the
immediate suspicion of cheating. Academic dishonesty often
stems from a complex interplay of factors including peer influences, parental
expectations, and institutional culture.
Moreover, as students progress through their academic journey, they encounter
varied challenges. While overt cheating might explain some anomalies
in academic data, the progression of subjects to more advanced levels could also
contribute to fluctuations in the distribution of grades [17]. However, this
influence might be relatively subtle within the broader context of academic
trends. This multifaceted view of academic integrity underscores the need for a
nuanced interpretation of data trends and student behavior.
6.4 Association Mining Rule
In the field of data mining and pattern recognition, association rule mining
stands as a fundamental technique for uncovering relationships and
dependencies within datasets. This analysis is particularly crucial in various
domains, including retail, marketing, and recommendation systems. In this
project, we delve into the exploration of association rules derived from
transactional data, aiming to uncover significant insights and patterns.
On this section, we aimed to answer the question of figuring out a way of
prediciting student’s cheating behavior. At the end, you will find that we did
answer that question, but that this research also managed to give us insights on
how students change their behavior during the semester.
6.5 Data Manipulation
In this study, clustering students based on their improvement or decline in
efficiency over the semester was conducted to better understand variations in
their academic performance. This methodology aimed to categorize students
into distinct groups based on their Efficiency Index—a metric representing the
relationship between time invested in homework and subsequent grade
improvement.
The process involved computing the Efficiency Index for each student based
on their homework completion times and the subsequent grade changes
achieved throughout the semester. The median Efficiency Index was calculated
for each student across their completed homework assignments, providing a
benchmark for their typical performance level. Students were then categorized
into groups:
Category A representing the highest 20 % (lowest efficiency), Category B for
the middle range, and Category C for the bottom 20% (highest efficiency).
The categorization process was dynamic, allowing students to move between
these categories based on changes in their Efficiency Index over subsequent
homework submissions. For instance, if a student consistently performed at
Category C but experienced a significant decline in efficiency, moving to
Category B, it signaled a deviation from their usual performance. Conversely, if
a student exhibited a sudden improvement, moving from Category B to Category
A, this was considered an unusual positive change in their performance
trajectory.
This clustering approach served as a red flag system, enabling the identification
of students whose performance patterns deviated substantially from
their established norms. For instance, a student maintaining a consistent low
efficiency level but displaying a sudden shift to a lower category at the same
time as unusually higher grades could be a cause for concern and prompt further
investigation by professors or academic staff. Such shifts in performance
categories indicated potential anomalies in academic behavior and could signify
a need for closer monitoring or support for the student in question.
6.6 Association Rule Analysis on Efficiency Index
6.6.1 Data Mining and Testing
The dataset under investigation contains a series of association rules, each
defined by antecedents and consequents, alongside corresponding metrics
indicative of the strength of association. Table 1 provides a summary of the key
association rules identified within the dataset.
6.6.2 Results
The table displays associations between different student categories (A, B, and
C) based on their Efficiency Index:
These findings suggest strong associations between certain categories. Notably,
there is a significant relationship between category B and category A
(very bad students) and vice versa, with relatively high confidence levels. Similarly,
there are distinct associations between categories B and C, and C and B,
indicating consistent patterns or transitions between these student performance
categories. These patterns may warrant further investigation or monitoring,
especially concerning shifts between categories that deviate from expected
student performance trends.
6.6.3 Interpretation of the Results
The transition probabilities between student categories, as indicated by the
association mining results, reveal intriguing insights into the dynamics of student
performance. The data suggests a relatively symmetrical transition between
categories, highlighting an approximately equal likelihood (41.3%) of movement
from a category denoting poor performance (bad student) to an average student,
and vice versa. This symmetrical shift suggests a considerable fluctuation in
student performance between these two categories.
Additionally, the analysis illustrates a 30.3% likelihood of an average student
transitioning to a good student category. This particular statistic is promising
as it signifies a favorable trend. When an average student progresses to a good
student category, there’s a higher probability that they will maintain this elevated
performance level. This pattern is significant as it hints at a degree
of consistency and stability in maintaining higher academic performance once
领英推荐
achieved. It implies that the efforts or interventions leading an average student
to improve towards becoming a good student may likely yield lasting positive
outcomes.
6.6.4 Conclusions
On the question we proposed to answer, which is :”How to predict cheating”,
we finally got our first good statistical answer. There is a very weak support
and confidence between A and C. It is highly unlikely that a consistently bad
student will become a good student out of nowhere, so if we are tracking the
students based on this category and we find out that this is happening, the
instructor can consider that the chances of cheating are very high.
Understanding these transition probabilities not only aids in identifying potential
trends in student performance but also provides valuable insights for educators
and institutions aiming to implement targeted strategies for academic
improvement. Identifying the crucial junctures where students tend to shift
categories helps in designing interventions to facilitate and sustain positive
progress in academic performance.
6.6.5 Limitations
Despite those findings, Calculating the standard deviation of the EI is a more
accurate method for spotting cheating compared to simply analyzing the
Efficiency Index alone. This is because the standard deviation captures the
variability in the Efficiency Index, thereby exposing potential outliers that might
be indicative of unethical practices.
For example, a student with a high Efficiency Index but a large standard
deviation suggests inconsistency in their performance and raises suspicion of
potential cheating. Conversely, a student with a moderate Efficiency Index but
a small standard deviation indicates consistent effort and genuine understanding
of the material. Therefore, analyzing both the Efficiency Index and its standard
deviation offers a more comprehensive picture of student performance and helps
identify potential cheating cases with greater accuracy.
6.7 Association Mining Analysis on Standard Deviation of EI
To mitigate the limitations above, I have decided that it would be clever to do
the same Association Mining Analysis with the Apriori algorithm, but instead
of using EI, I used the standard deviation of EI. This would decrease the risk
of False Positives on cheating and will provide less red flags, but more accuracy
on our predictions.
Furthermore, the Efficiency Index clustering style was based on comparing
the student with their own previous performance, providing a way of measuring
differences in their own behavior. Standard Deviation EI clustering will not
compare the student with himself, but will find erratic student behavior based
on his whole class. It is more broad and, used together with the EI clustering,
will provide a deeper understanding of their behavior.
6.7.1 Data Manipulation and Tests
To do so, I first deleted the outliers based on total standard deviation above
3. Then, I calculated the standard deviation of every student in a certain
homework. I classified students the same way as in the previous classification,
so it the 20% worse are A, 20% best are C and the rest are B. After that, in
every new assignment, new grades came and with this, new classifications.
6.7.2 Results
The explanation of the results is the same as before because we used the same
algorithm. The reader can look with more details at Table 2
6.7.3 Interpretation and Conclusion
The Results have shown to be very similar to the previous one, but as predicted,
with a higher accuracy. If a student is spotted going from A to C in a homework,
it is extremely likely that he did something that he was not supposed to.
7 Conclusion
In conclusion, we found out that procrastination is part of the life of everyone,
not excluding the students. It is not necessarily laziness and it is negatively
correlated to bad grades, thus answering the first question we have proposed
ourselves to answer. We discovered that students do learn with their mistakes
and tend to improve with time when they are punished, but not if the
punishment is too harsh. We also delved deeply into cheating behavior, found
out how and why students do it, as well as developed a new tool, called
Efficiency Index, to implement an algorithm that will spot if a student is cheating,
or at least give an important red flag.
Overall, this was a productive and deep paper on Student Behavior analysis
and provided information and tools necessary for helping instructors better
understand students. I hope this will help improving the world’s education.
?
References
[1] Ahn, J., and McEachin, A. Student enrollment patterns and achievement
in ohio’s online charter schools. Educational Researcher 46, 1 (2017),
44–57.
[2] Akc，ap?nar, G., and Kokoc，, M. Analyzing the relationship between
student’s assignment submission behaviors and course achievement through
process mining analysis. Turkish Journal of Computer and Mathematics
Education (TURCOMAT) (2020).
[3] Alhothali, A., Albsisi, M., Assalahi, H., and Aldosemani, T. Predicting
student outcomes in online courses using machine learning techniques:
A review. Sustainability 14 (2022), 6199.
[4] Ambrose, S. A., et al. How learning works: Seven research-based principles
for smart teaching. Jossey-Bass (2010).
[5] Anderman, E. M., and Murdock, T. B. Perceptions of ability, value,
and choice as predictors of academic cheating. Contemporary Educational
Psychology 25, 1 (2000), 21–37.
[6] Anderman, E. M., and Murdock, T. B. 1 - the psychology of academic
cheating. In Psychology of Academic Cheating, E. M. Anderman and T. B.
Murdock, Eds. Academic Press, Burlington, 2007, pp. 1–5.
[7] Anonymous. The effects of failing grades on student motivation and behavior.
Journal of Educational Psychology (2017).
21
[8] Butler, R., and Neuman, S. B. The impact of feedback on student
learning and motivation. Educational Psychology Review 24, 1 (2012), 61–
89.
[9] Chang, H., and Kim, H. Predicting the pass probability of secondary
school students taking online classes. Computers & Education 164 (2020),
104110.
[10] Chen, Wang, and Cheng. The relationship between online assignment
submission behavior and academic performance in higher education: A
systematic literature review.
[11] Cheng, S.-L., and Xie, K. Why college students procrastinate in online
courses: A self-regulated learning perspective. The Internet and Higher
Education 50 (2021), 100807.
[12] Chickering, A. W., and Gamson, Z. F. Seven principles for good
practice in undergraduate education. AAHE Bulletin 39, 7 (1987), 3–7.
[13] Falkner. Procrastination and delayed assignment submissions: Student
and faculty perceptions of late point policy and grace within an online
learning environment. Online Learning 24 (2012), 10.24059/olj.v24i3.2302.
[14] Gregory, K., and Mor′on-Garc′?a, S. Assignment submission, student
behavior and experience. Engineering Education 4, 1 (2009), 16–28.
[15] Heissel, J. The relative benefits of live versus online delivery: Evidence
from virtual algebra i in north carolina. Economics of Education Review
53 (2016), 99–115.
[16] Hooshyar, D., Pedaste, M., Wang, M., Huang, Y., and Lim, H.
Prediction of students’ procrastination behavior through their submission
behavioral pattern in online learning. Journal of Ambient Intelligence and
Humanized Computing (2020).
[17] Jones, E. Cheating in College: Why Students Do It and What Educators
Can Do About It. Johns Hopkins University Press, 2010.
[18] Klingsieck, K. B. Why students procrastinate: A qualitative approach.
Journal of College Student Development 54, 4 (2013), 397–412.
[19] Ko, S. Teaching online: A guide to best practices. Link.
[20] Kokoc，, A., and G¨ulc，er. Students’ online assignment submission behavior
and its relationship with end-of-term academic performance: A study
using learning analytics.
[21] Kuh, G. D. Student engagement in college: What professors do. College
Teaching 53, 3 (2005), 107–111.
22
[22] Laksitowening, K., Prasetya, M., Suwawi, D., and Herdiani, A.
Capturing students’ dynamic learning pattern based on activity logs using
hierarchical clustering. Jurnal RESTI (Rekayasa Sistem dan Teknologi
Informasi) 7 (2023), 34–40.
[23] Miron, G., and Urschel, J. Understanding and improving full-time
virtual schools: A study of student characteristics, school finance, and
school performance in schools operated by k12 inc. National Education
Policy Center (2012).
[24] Naceanceno, K., and Akpanudo, U. Learning environment and online
assignment behaviors as predictors of the academic performance of low
socioeconomic status (ses) junior high students. International Journal of
Technology in Education and Science 7 (2023), 253–273.
[25] Onwuegbuzie, A. J., and Daley, C. G. Anxiety, procrastination, and
academic performance among undergraduate students. Educational Research
and Evaluation 15, 4 (2009), 287–306.
[26] Pascarella, E. T., and Terenzini, P. T. The impact of first-generation
college students on college outcomes. Research in Higher Education 49, 7
(2008), 595–634.
[27] Pychyl, T. A. Procrastination isn’t laziness: It’s an emotion regulation
problem. Psychological Science 24, 3 (2013), 445–453.
[28] Santelli, B., Robertson, S., Larson, E., and Humphrey, S. Procrastination
and delayed assignment submissions: Student and faculty perceptions
of late point policy and grace. Online Learning 24 (2020).
[29] Schwen, K., and Lehman, C. Motivating online learners: Strategies for
student engagement and success. Link.
[30] Siemens, G. Learning analytics for online classrooms: A practical guide
for instructors. Link.
[31] Siemens, G., Gasevic, D., and Baker, R. Predicting student success
in online courses using learning analytics: A literature review.
[32] Sirois, F. M., and Pychyl, T. A. The procrastination equation: How
feelings of inadequacy and lack of control contribute to procrastination.
Journal of Personality and Social Psychology 104, 3 (2013), 346–364.
[33] Steel, P. The distinction between laziness and procrastination. Motivation
and Emotion 31, 3 (2007), 167–184.
[34] Steel, P. The procrastination equation: How to stop putting things off
and get things done. Link.
23
[35] Tan, T., and Samavedham, L. The learning process matters: A sequence
analysis perspective of examining procrastination using learning
management system. Computers and Education Open 3 (2022), 100112.
[36] Walker, L., and McCabe, D. L. Academic integrity across nations: An
international study of faculty and student attitudes and behaviors. Journal
of Education for Business 89, 2 (2014), 68–78.
[37] Wolters, C. A. The impact of self-efficacy on procrastination and late
assignment submission. Journal of Educational Psychology 103, 2 (2011),
337.