Modeling the Student Success or Failure in Engineering at VUT Using the Date Band Algorithm

Invention Journal of Research Technology in Engineering & Management (IJRTEM)
ISSN: 2455-3689
www.ijrtem.com Volume 2 Issue 4 ǁ April 2018 ǁ PP 48-55
|Volume 2| Issue 4 | www.ijrtem.com | 48 |
Modeling the Student Success or Failure in Engineering at VUT
Using the Date Band Algorithm
1,
Langa Hendrick Musawenkosi , 2,
Twala Bhekisipho2
1
Department of Electrical Engineering, Vaal University of Technology, Vanderbijlpark, South Africa
2
Department of Electrical & Mining Engineering, University of South Africa , Florida, Johannesburg, South
Africa
ABSTRACT: The success or failure of students is a concern for every academic institution, college, university,
governments and students themselves. This paper presents a model to determine the propensity of a student to
succeed in the Electrical Engineering Department at Vaal University of Technology. Firstly, various machine
learning algorithms which can be used in modelling and in predicting student success or failure are discussed as
well a new algorithm called the date band algorithm. Secondly, the concept of an academic model is also
discussed. This model defines the domain and focus of data used to make preditcions. The academic model consists
of the subject, the lecturer and the sudent each of which has various attributes. One of the attributes discussed in
this paper is the popularity index which is a measure of cohesiveness of the model.The Date Band Algorithm is
presented among others in the development of the model. In this algorithm, predictions are made to optimize the
performance of academic environment, thereby impacting on the choices of funders when they support students.
KEYWORDS: Academic Environment Model, Date Band Algorithm, Decision Trees, K-Nearest Neighbor,
Machine Learning;
I. INTRODUCTION
The success rate in academic environments is not only a concern for those institutions but also governments,
sponsors such as the public and the private sectors, parents, students themselves and other stakeholders. It therefore
behooves us to investigate the propensity of those students to succeed using scientific methods such as machine
learning. Machine Learning (ML) has a variety of algorithms that can be applied in addressing this problem.The
South African government and funders can save a lot of resources when funding these institutions. Therefore, the
application of rigorous methods of machine learning can improve the efficiency in the academic sector. For the
most part in South Africa, the largest contributor of funding in public education is government, that is, the ministry
of education. Although the ministry of education takes no account of income that is raised from student fees and
other private sources, these public institutions have to account by submitting annual financial statements which
reflect all income and all expenditure from all public and private sources [8].
The need to attract and retain students in engineering programs remains by necessity a focal point of interest and
effort in engineering education, [12]. All universities and colleges have marketing departments to make sure that
they attract the best of the best. They run various marketing programs for this purpose. Paul and Cowe Falls, [3]
highlighted the three aspects for engineering careers success based on the availability of the resources. Firstly,
lifelong learning is fundamental for success in the 21st
century engineering career. Staying abreast with the most
recent technological advancement is essential for being innovative and creative. Secondly a study in the
engineering construction industry is the most critical aspect of fostering a successful career path was in developing
a career network. This includes networking, mentorship training and constructive feedback. Thirdly, the aspect of
engineering career success relates to the models “proactive personality” variable.
II. MACHINE LEARNING ALGORITHMS OVERVIEW
Decision Trees, DTs : Decision Trees, DTs [12] are simple yet successful techniques for supervised classification
learning. This classification method consists of decision nodes, connected by branches, extending from the root
node until the terminating leaf nodes, [1]. Starting at the root node attributes are tested at the decision node, with
each possible outcome resulting in a branch. A decision tree algorithm aims to recursively split the observations
into mutually exclusive subgroups until there is not further split that makes a difference in terms of statistical or
impurity measures, [4]. A path is traced from the root to a leaf node which holds the class predication for that
sample. Decision trees can easily be converted into IF-THEN rules and used for decision making, [9].

Modeling the Student Success or Failure in Engineering…
The K-Nearest Neighbor Classifier, KNN : The K – Nearest Neighbor is an example of instance based
learning, in which the data set is stored, so that the classification for a new unclassified record may be found by
simply comparing it to the most similar records in the training set [5]. The KNN algorithm is the earliest researched
algorithm used for classification and is proved as one of the algorithms which have good classification results, but
there are still some problems that need to be attended to. For example, it is not yet settled, how to select the value
of k and how to select feature sets to make the classification better and their impact on each other [11].
The Support Vector Machine, SVM :Support Vector Machines (SVM) is an algorithm that uses nonlinear
mapping to transform the original data into a higher dimension [7]. SVM’s are pattern classifiers that can be
expressed in the form of hyper planes to discriminate between positive instances and negative instances pioneered
by Vapkin [10].
The Artificial Neural Network, ANN : The artificial neural network (ANN) is an information processing
paradigm that is inspired by the way biological nervous systems, like the brain process information [6]. ANN’s
are powerful tools that can be used to learn patterns from data. According to [5] a neural network is a collection
of nodes that are connected in some pattern to allow communication between the nodes. These nodes, also referred
to as neurons or units, are simple processors whose computing ability is typically restricted to a rule for combining
input signals and an activation rule that takes the combined input to calculate the output signal. Output signals
may be sent to other nodes along connections known as weights. The weights usually excite or inhibit the signal
that is being communicated.
Fig. 1. Artificial Neural Network
An example of neural network is shown in Fig. 1. Since the artificial neuron mimics the natural neuron, neural
networks therefore mimics the operation of the brain and can be quite useful in making classifications and
predictions.
III. MODELING AN ACADEMIC ENVIRONMENT
The Academic Environment Model : The environment is the sum total of surroundings of a living organism
including natural forces and other things, which provide conditions for development and growth as well as danger
and damage. An academic environment, in figure 2, is where a student exists. In order to model this, it is necessary
to gather information about the student, the lecturer and the module which will form part of the environment.
Input nodes layer Output nodes layer
Hidden nodes layer
Input x1
Input x2
Input x3
Output y1
Output y2
LoutLink
s

Fig. 2. Academic Environment Model
Tinto’s (1975) Student Integration Model (SIM) postulated that students who persist and succeed in college are
those who are able to integrate successfully into an institution’s social an academic environment. Alternatively,
the students who are more likely to struggle and fail to persist are those who do not attempt or achieve social and
academic integration.
The Lecturer’s Popularity Index : One of the common problems in higher education is the evaluation of the
instructor’s performances in a course. The most widely used tool to evaluate instructors’ performance in a course
is through surveying students’ responses about the course and its instructor through a questionnaire [2]. The
percentage score of likes for a given the subject or the lecturer is given by the following equation:
nk
x
xL
n
i
i
 1
)( (1)
Where:
n = number of instances of likes for a lecturer or subject and
k = the highest number of likes in an instance
Clearly, some lecturers are more popular than others. There are lecturers whom students really detest and there
are lecturers whom they adore. These can be due to several reasons, such as the appearance, teaching style, level
of education, leadership and so on. The popularity of the lecturer has a correlation with the performance of the
student. Similarly, the popularity of the subject can be measured.
The Academic Environment Client System : The panel below allows the student to rate the lecturer using
choices of numbers between 1 and 5 as shown in Fig 3. If the lecturer is least popular then the choice would be a
1 and if the lecturer is a student’s favorite then the choice would be a 5. This information is then captured in a
database for future references. The popularity of the lecturer can increase or decrease with time depending on the
performance of the lecturer. This is an important feature to have in the design.
Fig. 3. Academic Environment Client System

A survey of approximately 600 students was completed where eleven lecturers were evaluated and eight subjects
were also evaluated using the system as shown in the figure below. Upon analyzing results, interesting
observations were noted. A lecturer teaches an average of 120 students per semester.
Popularity Indices for Power Engineering Lecturers :It is clear from the graph that some lecturers are more
popular than others and that no lecturer obtained less than 50% which is fair enough in terms of the quality of
lecturers employed in the department of Power Engineering. There are various reasons why a lecturer would be
unpopular. It could just be shear laziness on his part, lack of understanding of the subject he teaches, the attitude,
very strict, to name a few but a few. And there are various reasons why a lecturer could be popular. It could be
that they are good in the subject matter, they have good qualifications, they are lenient, their attitude, again to
name a few.
Fig. 4. Popularity Indices for Power Engineering Lecturers
Evidently, as seen in Fig. 4, some lecturers are more popular than others in this department as would be the case
with other departments.
Throughputs for 2016 semester 1 & 2 : There are many ML algorithms used for making predictions. This paper
focuses on the algorithm called the date band algorithm with special reference to making predictions of student
success or failure. Some algorithms will be more accurate than others and some will be more appropriate than others
also. Evidently, as shown in Fig. 5, some subjects are more popular than others in this department as would be the
case with other departments.
Fig. 5. Throughput for Semester 1 & 2 in 2016
72.63
88.75
64.44
64.21
94.19
58.0058.82
68.00
75.45
69.23
59.20
0.00
20.00
40.00
60.00
80.00
100.00
Power Engineering Lecturers
Indices

The x axis is the representation of subject codes for the department of Power Engineering. T Put% S1 and T Put%
S1 are percentage throughputs for semester 1 and 2 respectively. An interesting observation in the graph is that
there is a general improvement in the throughput for all subjects except for EPPRT3C and a slight decrease in
EPDES3B. While it is plausible to observe an increase in the throughput it is conversely concerning to see a decline
in the throughput. This decline in the throughput, however, can be addressed with the lecturers who are responsible
for teaching those subjects. On the other hand, for those subjects that have seen improvement, either maintain the
position or strive for more improvement.
IV. THE DATE BAND ALGORITHM
The date band algorithm, shown in Algorithm 1.1, is derived from an assumption that there exists a relationship
between the date of birth of the student attribute and the propensity of that student to pass or fail. The algorithm
scans the date band to determine the number of students that succeed in that band and those that fail in that band.
To predict the probability of the student instance to succeed, it is a question of classifying the student according to
that probability.
Algorithm 1.1 Date Band Algorithm
Start
{
int Count = 0, Passed = 0, n = 0;
Enter date of birth
Read date of birth,
for ( n = 1, to number of entries, n++)
{
Scan the list of students for students within this date band
If (Found) then
{
Count++;
Check if Successful
If (Successful) then
{
Passed++
}
End if
}
End if
}
End for
Predict Probability for Success
}
End Start
So, this algorithm, as it has been stated, works on the premise that students can first be classified in date bands as
far as their date of birth are concerned and then a prediction of the probability of that student to succeed or fail in
the academic environment can be made. The following graph reveals some interesting results.

Fig. 6. Date Band AlgorithmProbabilities
Fig. 7.
It is clear from this bar graph, in Fig. 6, that Date band – 6 scores the highest, and this means that the students in
this band have 100 % probability that they will succeed in when enrolled for this subject. The lowest in this test
is Band – 12 with 37.5 % probability. Band – 7 had a 0 % probability, but the reason it is zero is that there were
no students in that band in fact.
V. RESULTS OF THE DATE BAND ALGORITHM
The academic environment consists of the student, the lecturer and the course or subject that the student enrolls
for. It has been shown that the condition of the environment can be measured by making use of the popularity
indices of lecturers and the courses themselves. So, when a student enters an environment, he or she will be
subjected to these conditions. On the other hand, the student has certain features which play a part in determining
his or her success. These features can include the APS score. The target variable could pass or fail, which implies
that we want to predict whether the student will pass or not given certain set of variables such as:
X1 = English
X2 = Mathematics
X3 = Physical Science
X4 = Students Birth Date Band
These variables were used also in the multiple regression model and the decision tree model to realize similar
results as shown in Fig. 7.
Fig. 8. Date Band Algorithm GUI
Other variables or features can be considered such a as age, place of birth, household income etc. and they would
have varying degrees of influence in the prediction outcome. For example, in the KNN algorithm, it has been found
that the most influential variables are Mathematics, Physics and Chemistry and English is least influential. In our
algorithm, we use the Date Band as an input variable although a combination with other features can be performed
as in a neural network for instance. A common standard of entry requirements for engineering schools in South
Africa are Mathematics, Science and English.

Fig. 9. Student Success Probalitiy Prediction Model Rerults
It is interesting in Fig. 8, to note that there is no data available for Band 7, therefore a prediction cannot be made.
This is because of the lack of historical data with which to make predictions. It is also that for Band 6 there is a
100% chance for success but it does not mean that, that position will necessarily remain the case because input
data sets change with time. When using the Date Band Algorithm, it has been found that a prediction can be made
on the basis of the environmental conditions that, it is known how the class of students are doing given a lecturer
and a subject and the most recent performance of that class. Consequently, it becomes a very dynamic model.
VI. CONCLUSION
This paper has presented the Date Band Algorithm as one of the solutions to predict the probability of a student
who wishes to enroll for a subject to succeed. So, the basic question that is asked by a student who wishes to take
a subject is this, “What are my chances of succeeding should I take a particular subject with lecturer so and so?”
He then enters that information in the system and the algorithm will generate a probability which can be essentially
regarded as an advice, a warning or information. It is in this paper, finally that the Date Band algorithm was used
to predict the propensity of the student who enters the academic environment in the department of Power
Engineering at the Vaal University of Technology.
REFERENCES
1. Larose, T.D., Larose, D.C.: Data Mining and Predictive Analytics, 2nd edn, Wiley, (2015).
2. Agaolgu, M.: Predicting Instructor Performance Using Data Mining Techniques in Higher Education, IEEE,
pp. 2379 – 2387, Turkey (2016).
3. Paul R.,Cowe Falls, L.: Mapping Career Success Competencies to Engieering Leadership Capabilities, IEEE,
pp. 1 – 6, Galgary, Canada (2015).
4. Agaolgu, M.: Predicting Instructor Performance Using Data Mining Techniques in Higher Education, IEEE
(2016).
5. Callan, R: Artificial Intelligence, Palgrave Macmillan (2003).
6. Abdella, M., Marwala, T.: Treatment of missing data using neural networks and genetic algorithms.
Proceedings of the International Joint Conference on Neural Networks, July 31 – August 23, pp. 598 – 603,
Montreal, Canada (2005).
7. Han, J., Kamber, M., Pei, J.: Data Mining 3rd edn, Morgan Kaufmann Publications (2012).
8. Ministry of Education. A New Funding Framework: How Government grants are allocated to Higher
Education Public Institutions. (February 2004).
9. Kamber, M., Winstone, L., Gong, W., Cheng, S., Han, J.: Generalization and Decision Tree Induction:
Efficient Classification in Data Mining, IEEE, pp. 111 – 120, (1997).
10. Twala, B.: Robot Execution Failure Prediction Using Incomplete Data. Proceedings of the 2009 IEEE
International Conference on Robotics and Biometrics, December pp. 1518 – 1523, China (2009).

11. Shang, W., Zhu, H.: The Improved Ontology kNN Algorithm and its Application, pp. 198 – 203, IEEE,
(2006).
12. Imbrie, P.K., Lin, J.: Work in Progress Engineering Students Change in Profile Over the Freshman Year
across Male and Female Samples: A Neural Network Approach, 36th
ASEE / IEEE Frontiers in Education
Conference, pp. 27 – 28, San Diego, CA (2006).

Modeling the Student Success or Failure in Engineering at VUT Using the Date Band Algorithm

More Related Content

What's hot

Similar to Modeling the Student Success or Failure in Engineering at VUT Using the Date Band Algorithm

More from journal ijrtem

Recently uploaded

Modeling the Student Success or Failure in Engineering at VUT Using the Date Band Algorithm