Analysis of student’s mark in fit faculty over 7-year period (2014-2020)

ANALYSIS OF STUDENT’S MARK IN FIT FACULTY OVER 7-

YEAR PERIOD (2014-2020)

SVTH: Vũ Hoàng Việt, Nguyễn Thế Đạt,

Tô n Nữ Tú Anh, Nguyễn Thị Hà My

GVHD: ThS Bùi Quốc Khá nh

Tó m tắt- Điểm số chí nh là một trong những yếu tố phản ảnh rõ nhất năng lực của từng sinh

viên của một mô n học. Việc tổng hợp và phân tích điểm số là một bước khô ng thể thiếu để khi nghiên

cứu về tì nh trạng học tập thực tế trong hiện tại và quá khứ, từ đó đưa ra dự bá o về tương lai hoặc đề

xuất giải phá p cho những vấn đề thực tiễn. Phâ n tí ch dữ liệu thô và minh họa nó ra biểu đồ là một

cá ch làm cho những số liệu khó hiểu trở nên gần gũi với người đọc. Do đó, bài viết sẽ đề cập và miêu

tả toàn bộ dữ liệu, thư viện, phương pháp, thuật toán được sử dụng để phá t triển dự á n. Tó m lại, bài

viết rút ra kết luận rằng mục đích của bài nghiên cứu này là đem lại những gó c nhì n mới trong việc

tiếp cận dữ liệu ở ngay khu vực mì nh tiếp xúc thực tế, mà đối tượng cụ thể trong bài nghiên cứu này

là các đầu điểm của khoa CNTT HANU và trì nh bày những lời khuyên cho những bạn sinh viên để có

thể nâ ng cao khả năng hoàn thành môn học mà cá c bạn tham gia.

Abstract - Score is one of the factors that best reflect the ability of each student’s performance

in a subject. The synthesis and analysis of scores is an indispensable step when studying the actual

learning performance situation in the present and the past, thereby making predictions about the

future or proposing solutions to practical problems. Analyzing the raw data and illustrating it

graphically is one way of making confusing data more accessible to the reader. Therefore, the article

will cover and describe the data, libraries, methods, and algorithms used to develop the project. In a

nutshell, the article concludes that the purpose of this research is to bring new perspectives to the

data at the very area where they are practically relevant to ourselves, specifically in this research, the

students’ scores in IT Faculty of HANU, and offer advice to students to improve their ability to

complete the subject they have participated.

I. INTRODUCTION

Normally, students can find scores with a basic statistical perspective, Excel files or pdf

files of scores are regularly uploaded to the FIT Portal informing name and how much scores

one has got, so they will only approach the surface data, limited information and other

limitations. The first one is about the relationships between the different data components. For

example, students know who has decreased or increased from the Midterm Exam to the Final

Exam but cannot know how many people have increased or decreased, account to what

percentage, with how much the score increases or decreases. The second limitation is about

the relationship between the distinctive score files, students can know all the scores

information for anyone taking any subject, any semester, but cannot figure out any

comparative connection between them is, often, people develop highly subjective perceptions.

92

These statements have an uncertain logical basis so they can be easily ignored in situations of

gain or loss, or they may be wrong and lead to risky action. higher and yield undesirable

results.

The most concern of students besides the knowledge and ability gained after the

courses, it is the completion of the subject. Course completion is of great significance to a

student as it plays a role in demonstrating a student's receptive level as well as in the rate at

which the program is completed. A student problem that is frequently discussed and

concerned before and during a course, is the failure to complete the course. This issue will

naturally continue due to the diversity of knowledge, the breadth and depth of the courses, or

differences in background levels or their own interests. It will be too simplified if you only

give the advice stating that, study diligently and try to study well (apply what you have learnt,

do it yourself, understand the lesson, actively ask and do not hesitate to ask questions, hard

work, ...) and when the ability comes from yourself, your score will automatically get high,

and you will pass the subject automatically. Oversimplification leads us away from truthful

understanding toward a distorted view of things, instead, having to step back and see the

bigger picture and learning how to connect the different dots to make sense of reality, is one

of the fundamental approach of this research [3].

Therefore, this research paper uses a different approach to data research and more

advanced information comprehension, which is Data Mining. Typically, data mining uncovers

interesting patterns and relationships hidden in a large volume of raw data, and the results

tapped out may help make valuable predictions or future observations in the real world [2].

What differentiates data science from statistic is that data science is a holistic approach

[1]. With the advantage of this method, the paper not only provides many perspectives on the

FIT test score dataset in many different angles and relationships, but most importantly, it

creates a theoretical basis, firmly give advice to students to be able to fully trust and thereby

reduce their chance not to complete the subject.

II. IMPLEMENTATION AND METHODOLOGIES

1. Data collection:

The research collected data on the FIT Portal. Especially in the two sections

Announcement and Course Mark Collection. There are two main types of files: Excel and

PDF. However, the research cannot collect all of the score files due to the limitation of data

sources. Despite of that, the research have collected sufficient amount of data for the analysis.

2. Data Preprocess

After collecting information from the FIT Portal, as a result of the fact that there are

many score files that each score file has a different format and different information, the

93

research have to filter out and only select information that is common and consistent among

all score files have such as “semester”, “subject”, “id”, “name”, “class” and “scores”

(Attendance score (10%), Midterm Exam score (30%/40%) and Final Exam score (60%).

Once the research have gathered enough information, the research start to convert all files

with only necessary information into Comma-separated Values (CSV) files to match the

code's algorithm. The cleaning involves changing special characters (k,b,#N/A,…), fills in -2

if the cell is left unprovided, -1 if the student did not attend, or drop rows, delete blank lines,

correct content etc. This process happens before and along with Implementing process

because this is a manual process and tiny errors in dataset can only be detected by

programming.

3. Implementation

The dataset prepared and collected is different from those available from data sources

on machine learning research organizations to test models and accuracy. Specifically, the

dataset doesn't have a "features - label" model commonly found in the field of machine

learning. This means that the research paper will not involve training model, instead, the goal

determined from scratch is to visualize the dataset and gain new perspectives from the score

charts along with other correlations.

The cleaned csv files from the raw files will be arranged in a tree directory structure,

with the score files of different subjects but for the same semester will be stored in the folder

named for that semester, the dataset will be aggregated of the 12-semester separated from the

original dataset and in each semester there are at most 17 csv files (score files) at least 4 files.

Along with the data storage structure of the directory tree, the algorithm will mainly use

the for loop function to access each folder ("ky") and the files containing the score ("mon")

according to the format of the path defined clearly.

94

The main part of the implementation consists of two steps, the first is to prepare the lists

needed for visualization, and the second is to use the matplotlib library to draw graphs from

those lists.

Depending on the types and characteristics of the plot to be drawn, the lists will be

predefined and become the target to be completed in step one. Normally, the lists will be

initialized in the form of a list containing only 0, and then after the loop to calculate, the

algorithm will automatically update the list and complete them after this first phase ends. One

of the most common lists is the "horizontal bar," which defines the display on the horizontal

x-axis, usually the name of the subject, the score or the timeline. The length of the "horizontal

bar" will be used to create other lists for more convenient. The study stipulates how to round

the score as follows: 4.51 => 5 <= 5.5, 5.51 => 6 <= 6.5, 6.51 => 7, etc. Along with that,

students who complete the course must have an overall score greater than or equal to 4.9.

Manipulating the matplotlib and NumPy libraries is the primary and important activity

of drawing charts from pre-prepared lists. The details of the diagram are quite diverse, from

the axes, labels, annotation as well as the aesthetic elements. Column charts are the most used

chart type for this study, because they emphasize the comparative factor between data and

also show the variation of the object, here is mainly scores. This part acts as the user interface.

Therefore, requirements for content, images, fonts, colors and proportions are carefully met.

This research paper has implemented 8 different types of plot types, the plots will be

saved according to that type of chart. With the normal running speed of an average computer,

the algorithm takes about less than 1 minute to generate about 150 images.

As mentioned earlier due to the specificity of datasets that have been manually collected

from the FIT Portal, this article does not have the training-related phases of the machine

learning studies with models and models evaluation criteria. Hence, the accuracy of the plots

is checked manually. After testing and fixing the errors in the algorithm and dataset, at the

end, the study has not encountered any cases of plot errors.

III. DISCUSSION AND FINDINGS

1. To increase their own course completion rates, students need to maximize their

Attendance scores

Data show that, although only accounting for 10% of the total score, attendance score

significantly affects course completion. Attendance Score, which is a scale that measures a

student's frequency of participation in a course, is usually measured according to the

frequency of classes, or the student's completion of homework. Attendance score only

accounts for 10% of the summary score, so the psychology of not taking this point seriously is

reasonable, a common manifestation in some students is willing to take a break or not to

95

submit assignments up to the limit of giving allowed. They accept a lower Attendance score

that remain within the allowable range and believe that the Midterm Exam and Final Exam

scores are the only determinant of whether or not they complete the course, and moreover, the

Attendance score is difficult to classify qualified and unqualified students. However, the

dataset has shown the opposite.

Students who complete a course always tend to have higher Attendance scores than

students who do not complete the course. Usually, Attendance scores are proportional to the

rate of course completion. In other words, the higher the Attendance score is, the higher the

percentage of students completing the course over the total number of students at that

Attendance score, (the more times the blue column (qualified) will be the higher the red

column (unqualified).

Looking at the picture (Attendance score of DSA in 2020 Spring), the feature is shown

quite clearly. Specifically, if there are 4 students with 0 Attendance score, all 4 students do

not complete DSA, the completion rate at this point is 0%; meanwhile, the 6-point completion

rate is definitely below 50% (the blue bar is lower than the red one); at a score of 7 the

increase is 50% (the two columns are equally high); At 8 points, it is definitely over 50% (the

blue bar is higher) and at both 9 and 10, 100% of students complete the course (there is no

case of students not completing the DSA at the levels this point). Thus, the percentage of

students completing the course tends to increase with the score of Attendance they receive,

although the Attendance score only accounts for 10% of the Final score. Similarly, in the

other data visualized, the same characteristic is evident in many other subjects, on different

semesters as well as in different shapes:

96

While there are subjects that exhibit a tendency for the percentage of completion to

increase with the Attendance score in the absolute way, most other subjects exhibit this

relative, sometimes, student completion rate of Attendance 10 is much lower than this at 9,

(eg. 2018 Spring Attendance PPL score). This is indicated by a large difference in height

between the red and blue bar at 10 Attendance score. If we calculate the details from the

chart, the percentage of students who completed and scored at 9 Attendance (15.4%) is

slightly higher with this rate at 10 (14.3%).

However, as we can see, all the students who complete the PPL course score a high

score of 8 or above, and all students with an Attendance score of 7 or below do not complete

the course (technically, as long as you score 7 Attendance, if you take one point lightly, it

97

only accounts for 10% of the Summary score, in PPL 2018 Spring, you will definitely not

complete the subject 100%). This shows that the Attendance score still maintains the

influence of the student's completion rate classification, which may be correct in some

subjects, while others are relative, the research paper. Research suggests that this may be due

to the influence of the Midterm Exam and the Final Exam.

The above hypothesis is partly confirmed, because the characteristics of the

proportionality between the Final Exam or the Midterm Exam scores, and the completion rate

of the course are shown much more clearly than the Attendance score. That is, in all the

available research data, the number of graphs showing this trend at the beginning of the Final

point is absolutely more absolutely correct than that of Midterm and many times more than

Attendance. This is easy to understand because Final Exam accounts for 60% and Midterm

Exam takes 30% 40% of the Final Score.

The ratio of Attendance scores to the rate of course completion leads to the

classification of students as well as the conclusion that, in order to increase their own

completion rate, students need to maximize their grades. Attendance among other points, and

this tip should not be taken lightly, even though it only accounts for 10% of the Summary

score.

2. When receiving Midterm results, students should make effort to improve their scores

in the Final Exam instead of unnecessary worry.

Understanding the importance of Final Exam scores to Final scores and course

completion rates is one thing anyone can realize, while as making effort to improve Final

scores after receiving Midterm results is a completely different thing due to the impact of

psychology inherent in some students. A type of score adds up to 60% of the Final score is no

one to dispute its impact on course completion, but it will also be a problem for students who

get Midterm scores that are not up to their expectations. , when they have the belief that Final

Exam scores will happen with the same situation. To illustrate this, for example, a student

who took DBS 2020 Fall and got 3 Midterm, after a while of calculating, he realized that he

needed more than 5 points to be able to complete the subject and on 6 points to make it

certain. But 3 points is quite a modest score compared to most of the participating students

somewhat demonstrating his current abilities, and the gap between completion and

incompletion forces you to increase the score from Midterm to Final is 2 - 3 points. If the

98

student finds this impossible to his or her ability, it is very difficult to try to improve his score

in this case.

The graph shows that about 50% of the students with a 3 will complete the course, so

that shows that 50% of them improved their score by 2-3 points, something the student is still

wondering about. after receiving Midterm points.

The data shows that, if the general ground had a clear increase in points from Midterm

to the Final, the majority would normally go up 1-3 points, the minority would increase 4-5,

and very little would increase by 6 or more. (eg. Midterm DBS increase or decrease over a

number of semesters)

99

Although it is possible that there will always be cases that contradict the situations in

the predictions, students still have evidence to believe in good and reasonable scenarios, that

once there is an upward trend, someone will definitely go up and go 1-3 points or more, and if

so, all the justifications for my starting point are baseless, and instead focus on improving the

score with action. and your beliefs are a wiser move.

3. Students should not be subjective to aim low for the Final Exam, once they know the

Midterm Exam scores.

The data shows that, from the 5 point or lower of the Final Exam, the failure rate will be

higher, but this is a milestone that some students consider it as the goal to strive for. The

research paper calls that behavior “aim low”.

Suppose in “aim low” psychology, a student got 6 Midterm in DMA in 2020 Fall, and

he did not go to school enough as encouraged, leading to Attendance score only 6 points.

According to the student's calculations, he only needs to take the Final Exam and get a little

over 4 (about 4.4) to make sure you will complete the course, along with that, he believes that

his Final Exam score cannot below average, because Midterm you were above average (5

points). So leading to the fact that he didn't really put in a lot of effort to prepare for the Final

Exam, he didn't have above average (6+) as his goal, or in other words, this student aim low.

Finally, the student gets 4, a score that is completely grounded and reasonable with his poor

preparation and you only anticipate when the questions in front of his are the exam questions.

This assumption can be somewhat tested based on the distribution of students who will

complete the course based on the 2020 Fall DMA Midterm Exam scores. With a score of 6, a

point above average, about a third of the students who scored this point (the height of the red

100

column is half the height of the blue bar) did not complete the course, most likely stemming

from psychology subjective and aim low

The data has shown that one of the main reasons the Final Exam is a top determinant

point besides it takes up 60% of the overall score, which is generally, this score is likely to be.

Student classification (pass, fail) is huge and much better than the Midterm Exam. This is

reflected in the difference between the boundary of the "completed" and "unfinished" area,

the more overlap or the higher the column length, the lower the classification ability.

IV CONCLUSION

Generally speaking, analysis of student's mark in FIT faculty over seven-year period

shows an overview of student learning. In order to get a higher passing rate and confidently

complete the subject, the students should try their best to complete their homework and not

absent to optimize their Attendance, also be well prepare for the Final Exam at any situation.

The research can be applied to further retrieve and analysis the complete database of FIT

faculty score in variety of aspects other than only score, it could be related to nationality, age

or gender.

101

REFERENCES

[1] M. Loukides, What Is Data Science? O’Reilly Media, 2011.

[2] D. Che, M. Safran, and Z. Peng, “From big data to big data mining: Challenges,

issues, and opportunities,” in Database Systems for Advanced Applications, vol. 1, Berlin,

Heidelberg: Springer Berlin Heidelberg, 2013, pp. 1–15.

[3] P. S. Powell, “The dangers of oversimplification,” Jubilee-centre.org. [Online].

Available: https://www.jubilee-centre.org/blog/the-dangers-of-oversimplification. [Accessed:

08-May-2021].

102