College of Banking and Financial Studies Data Warehousing Calculation Homework (Code: A92/ date 27/6/2029) ************* Which subject: Data warehous

College of Banking and Financial Studies Data Warehousing Calculation Homework (Code: A92/ date 27/6/2029)

*************

Which subject:

Data warehouses

If there is any program touse :

Oracle

Calculations or thury:

thury

How many words:

12 pages( 1800-2000word)

***********************
turnintin report needed

Similarity less than 5%

use : Oracle pro Database Systems Journal, vol. X/2019
33
Factors that contribute programming skill and CGPA as a CS graduate:
Mining Educational Data
Md Aref BILLAH, Sheikh Arif AHMED, Shahidul Islam KHAN
Department of Computer Science and Engineering (CSE)
International Islamic University Chittagong (IIUC)
Chittagong, Bangladesh
mdarefbillah818@gmail.com, sheikharif1993@gmail.com, nayeemkh@gmail.com
Computer Science (CS) has become one of the most popular under graduate program in last
few years. According to UGC roughly 116 universities out of 136 are offering computer
science program which indicates a massive number of students are choosing this program as
their undergraduate program. But statistically significant number of students are failing to
become skilled and effective CS graduate because many students are taking CS without
accessing their chance in this program. Success in academic and professional life require to
choose right under graduate program. Considering CGPA and Programming Skill as two of
the most significant factors to determine student’s success in CS, we have predicted these two
by taking students personal interest, academic results, analytical skill and problem solving
skill into account. We also extracted most significant features of a prospective CS student by
using gain ratio.
Keywords: Computer Science Student, Predicting Performance, Machine Learning
Techniques, Data Mining, Programming skill, CGPA
1
Introduction
Computer Science has now become a
buzzword in the global community. Being
one of the developing countries Bangladesh
Government has already taken the challenge
of outshining in the ICT department, so as
the students. But when the question of skill,
show casing talent and achievements in
national and international level comes it
seems significantly important percent of
students are failing to do so. Without proper
analysis, substantial amount of students are
taking this program and eventually the
performance of larger part of students of this
Program is performing poorly which is
hurting their
Academic and professional life. Though
massive number of students are rushing into
this program in Bangladesh, many IT
industries are still hiring IT professionals
from India because of limited number of
skilled graduates [1].
So, to predicting the performance of
prospective CS graduates before they start is
what they need. If they can know the factors
on what their performance as a CS graduate
depends, they can decide whether they are
going to take CS or not. There is a
possibility that they can change themselves
as the demands to be a good CS graduate.
In recent past, Students’ final result
predictor and Classification model for
determining students’ future was built by
taking different types of feature sets like as
previous academic result, family income,
family expenses, medium of teaching,
marital status, parent’s occupation, parent’s
qualification, family size, attendance,
assignment, lab work: [2],[3],[4],[5],[6],[7].
Different types of classification model were
built by implementing Decision tree,
Support Vector Machine and Naive Bayes
Classifier algorithm on students’ academic
results: [2], [3].
Purpose of this work is to build a model to
predict students’ final result or we can say
CGPA of their CS program and also the
programming skill which is very good
indicator of a good CS graduate before they
start.
To predict students’ performance in any
program it is important to take the previous
34
Factors that contribute programming skill and CGPA as a CS graduate: Mining Educational Data
academic results into account [8]. Studies
show that CS students’ need mathematical
skill for increasing programming and other
skills like problem solving and analytical
skills: [9], [10].
So besides academic
results, to make our proposed model more
accurate in this paper we used their previous
academic results, their experience with ICT
course at HSC level, online class experience,
Internet browsing reasons, Participation in
Mathematics or Science Olympiad, Interest
in Competitive programming and students
problem solving skills etc. To train our
model we have used current CS student’s
data and when our model is ready we
predicted both CGPA and programming
skill. Where CGPA tells about student’s
theoretical knowledge and programming
skill tells about their practical knowledge.
2 Literature Review
Many researcher worked for predicting the
performance of students.
Romero et al. (2013) conducted a study on
students participating in on-line discussion
forum and predicted students’ final
performance in Spain. They collected forum
interaction data such as number of messages
post/read, ask and reply relationship between
students. Afterwards compared them in
between classification and classification via
clustering approach. [11]
Alharbi et al. (2016) did a case study
whether they can highlight performance
problems early on and propose remedial
actions using data mining techniques. They
collected students data during admission and
after completing their academic first year
and eventually predicted good honors
outcomes with reasonable accuracy by using
classifying model with highlighting students
that are predicted to low achievers with high
probability module results [12].
Baradwaj & Pal (2012) suggested a
classification
model
for
Predicting
Performance
improvement
on
the
Educational databases which contains
invisible information for improvement of
students’ performance. They collected
300(74 females, 226 males) students record
from Dr. R. M. L. Awadh University,
Fazibad India and used Bayes classification
approach. They further conducted a research
by taking students class test, assignments,
attendance, lab work and seminars into
consideration and analyzed students’
performance in the semester final
examination. They used decision tree for
predicting students’ performance: [2], [3].
Daud et al. (2017) conducted a study where
using Educational Data mining approach
they
considered
student’s
family
expenditure, personal information and
predicted whether he will be able to
complete his degree or not. They used
WEKA tool to classify 100 students record
with 23 features each from different
universities of Pakistan [4]
Goga et al. (2015) proposed a tool by using
.Net framework which takes students various
information as input and predicts students’
grade. They first collected the student’s
enrollment
records
from
Babcock
University, Nigeria and then built models
using classification trees and a multi-layer
perception learning algorithms operating on
WEKA. In the domain of this study random
tree adopted as the best algorithm and served
as a building block of this generic system
[13]
Arsad & Buniyamin (2013) conducted a
study to predict student’s final result at the
Faculty of Electrical Engineering, Universiti
Teknologi MARA (UiTM), Malaysia. They
showed Artificial Neural Network model
served as a vital to predict students result.
They found that Students’ first and thirds
semesters fundamental course results
reflected their final result. Therefore,
fundamental subjects should be fully
understood because without clearing
fundamental knowledge it is very difficult to
face advance courses [14].
Huang & Fang (2012) proposed a
mathematical model for making early
prediction of students’ final score and
Engineering dynamic courses using the
results of first semester and validated using
next three semester data. They collected data
from 1900 engineering students of four
Database Systems Journal, vol. X/2019
semesters. Four different mathematical
modeling approaches: radial basis function
neural networks, multi-layer perception
neural
networks,
multivariate
linear
regression, multi-layer perception neural
networks, and support vector machines were
applied and experimental results showed
anyone can be used to predict the desired
result [15].
3 Techniques
We have used few data miming techniques
and existing algorithms and tested their
efficiency through various techniques. Here
is a short description of these tools and
techniques
The term Data mining is a misnomer, as it
refers to extracting knowledge from large
amount of data [16]. The data mining
techniques is featured to create model which
will help to find new data using unknown
data [17]. Data mining can be basically of
two types, Predictive and Descriptive [18].
Predictive Data Mining Model
This method studies previous historical data
and predict and forecast what is going to
happen to future data set e.g.: classification,
regression, time series analysis etc. [19].
C 4.5: It’s a decision tree based algorithm
for classification both numeric and nominal
classes. It was written by J. R. Quinlan [20].
Support Vector Machine: Support vector
machine is the supervised algorithm for
using as a classifier or regression algorithm
for pattern, nested problems or mining of
texts [21]. It uses a hyper plane to partition
two different classes where support vectors
are those which instances are used as the
margin [22].
Descriptive Data Mining Model
Descriptive data model is a predictive model
consisting of clustering, summarization,
association rule etc. It finds pattern in large
data and further works in intelligence system
decision-making.
Data mining is of two types according to its
class i.e. supervised and unsupervised
learning algorithms.
Unsupervised algorithms: No information
about class or class label. E.g.: Clustering
and association rules etc.
35
Supervised algorithms: Class data is known.
E.g.: classification, regression etc.
4 Methodology
Researchers from different regions found
that study on forecasting student’s final
result or CGPA for under graduation
program is very important. They considered
students’ academic result, parents data,
hours spent in study, activity in online
discussion forum, marital status & student’s
class attendance for forecasting their final
performance and building classification
model which can help them take decision
which track they should follow for the under
graduation program. Different from the
literature, we considered only Computer
Science program as our purpose of study for
its immeasurable popularity and the kind of
challenges it encounters. Especially, for
someone who is not ready to take those
challenges is sure to suffocate in CS. For CS
programming skill measurement is also very
important. So we build a model to predict
CGPA and programming skill using
different regression algorithm to build a
prediction model based on 23 features
including their academic results, experience
with ICT course, personal interests, personal
experiences of current students’ with similar
academic achievements and problem solving
skills.
Our proposed method has few stages1. Data collection
2. Pre-processing
•
Data cleaning
•
Transformation
•
Integration
•
Standardization
•
Feature selection
3. Data mining, model generation and
Performance
measurement
of
algorithms
4. Finally we will get a model to use.
36
Factors that contribute programming skill and CGPA as a CS graduate: Mining Educational Data
Fig. 1. Overview of Model. It shows the
whole process of building our model step by
step. Starts with data collection and ends
with building a model.
5 Implementation and Result
Primary data was collected by doing a
survey on current students of Computer
Science. We collected CS student’s data
having 23 features in these six following
categories. i.e.: Personal information,
personal Interest, academic results, their
experience with ICT course in HSC level,
level of their problem solving skill and their
online course experience. We assigned
numeric values to these questions responses
for data analysis and research purpose. We
have removed inconsistency from our data
set and transformed them into the proper
format to use them for analysis. Table 1
Shows the options with correspondent
values and short term of the survey
questionnaire for students and here is the
questionnaireSection A- Personal Information
A1- What is your name?
A2- Gender?
A3 -What is the Location of your
University?
Section B- Academic result
B1- What is your SSC GPA out of 5.00?
B2- What is your HSC GPA?
Section C- Experiences with ICT Course
C1- How did you find ICT course (Your
experience with ICT)?
C2- How was your academic result in ICT?
Section D- Personal Interest
D1-Why do you browse internet mostly?
D2- Participated in number of math or
science Olympiad?
D3- Online courses you’ve followed related
to your study?
D4- Rate your interest in Competitive
Programming?
Section E- Personal Experience
E1- “Tendency of using online resources
help a lot in Computer Science”:
E2- “Having Patience help in Computer
Science”:
E3- “Knowing PC configuration helps in
reading Computer Science”:
E4- ‘Computer gamer have better chance in
CS’:
E5- “Capability of Self-study makes
significant difference in Computer Science”:
E6- “HSC ICT Course result reflects once
potential in Computer Science”:
E7- Knowledge (about the CS program) you
had before starting your Program?
Section F- Problem solving skills and
others
F1- Rate your programming skill:
F2- Your skill in Basic mathematics (SSC
level):
F3- Your skill in Higher Level Mathematics
(HSC level):
F4- Rate your patience out of 10:
F5- Rate your capability of self-study out of
10:
Table 1. Options with correspondence values and short terms
Options
Short Term
Strongly Agree/Very Good
SA/VG/C3.8AA/5P/MT/VI/9/5:00P/Many /Very Interesting /9
10R/FLP/SMAE
to 10 ratings/Social media and
Entertainment
Values
5
Database Systems Journal, vol. X/2019
Agree/Good /4.50P and
above/More than 5/Interesting/7
to 8 ratings/Study and Social
Media
Neutral/Satisfactory/ /4.00P and
above/More than Once/5-6
ratings/, Reading Blogs only
Disagree/Less than satisfactory
/3.50P and above/Once/3-4
ratings/Online Courses only
Strongly disagree/Poor/ Less than
3.50P/Never/1-2 ratings/Online
Course and Reading Blogs
37
A/G/C3.4AA/4.5PAA/MT5/I/78R
4
N/S/C3.0AA/4.0PAA/MTO/56R
3
D/LTS/C2.5 AA/3.5PAA/O/34R
2
SD/P/CLT2.5/LT3.5P/Ne/1-2R
1
Table 2. Students related variables that illustrates the questions we asked the students and
probable answers.
Variables
Description
Possible Values
Gender
Students Gender
{Male, Female}
UniLoc
University Location
{Dhaka, Chittagong, Other City,
Outside City}
GSSC
Students grade in SSC
{5.00
>4.50 and 4.00 and 3.5 and 3.5}
GHSC
Students grade in HSC
{5.00
>4.50 and 4.00 and 3.5 and 3.5}
ICTResult Students ICT result in HSC {Very Good, Good, Satisfactory, Less
than satisfactory, poor}
ProRatings Students programming skill
{ 9 to 10,7 to 8,
ratings
5 to 6,3 to 4,1 to 2}
Bmath
Students skill in basic
{Very Good, Good, Satisfactory, Less
mathematics
than satisfactory, poor}
HMath
PaRatings
ICTExp
IBReason
MSOlym
Students skill in Higher
level mathematic
Students Patience’s ratings
Students experience with
ICT
Most important reason
behind browsing internet
Student participation in
number of math or science
{Very Good, Good, Satisfactory, Less
than satisfactory, poor}
{ 9 to 10,7 to 8,
5 to 6,3 to 4,1 to 2}
{Very Interesting, Interesting, Neutral,
Difficult, Very Difficult }
{Online courses and Reading Blogs,
Online Courses only, Reading Blogs
only, Study and Social Media, Social
media and Entertainment}
{many , less than 5, more than once,
once, never}
38
Factors that contribute programming skill and CGPA as a CS graduate: Mining Educational Data
PH
Olympiad
Number of online courses
followed
Students competitive
programming Interest ratings
Students capability of selfstudy ratings
Tendency of using online
resources helps
Having Patience helps
SSH
Self-study helps
CGC
Computer gamers chance in
Computer Science
ICT result reflects success in
CSE
Knowing about PC
configuration helps in
reading Computer Science?
NOC
Cpro
SSR
ORH
ICTRR
PC
We have collected data from current CS
student of various university from various
city of Bangladesh. Table 3 shows the
{many, less than 5, more than once,
once, never}
{ 9 to 10,7to 8,
5 to 6,3to 4,1 to 2}
{ 9 to 10,7to 8,
5 to 6, 3to 4,1 to 2}
{Strongly agree, agree, neutral,
disagree, strongly disagree}
{Strongly agree, agree, neutral,
disagree, strongly disagree}
{Strongly agree, agree, neutral,
disagree, strongly disagree}
{Strongly agree, agree, neutral,
disagree, strongly disagree}
{Strongly agree, agree, neutral,
disagree, strongly disagree}
{Strongly agree, agree, neutral,
disagree, strongly disagree}
frequency, percent valid percent and
cumulative percent of dataset in respect to
Gender
Table 3. Frequency Table (Gender of Students). 329 of 501 students are male and 172 is
female. That is 65.7% students are male and 34.3% students are female.
Frequency
Valid
Male
Female
Total
329
172
501
Percent
Valid Percent
65.7
34.3
100.0
65.7
34.3
100.0
Cumulative
Percent
65.7
100.0
Table 4. Programming skill * CGPA cross tabulation. We can see most of the good
programmers (with rating 4 and 5) have CGPA 3.5-4.00 and out of 90 students with
programming skill 5, 72 students have CGPA 3.7-4.0.
Programming skill * CGPA Cross tabulation
CGPA
2.492.5bellow
2.99
Programming skill
1
5
9
2
0
36
3
0
15
4
0
0
5
1
2
Total
6
62
Total
3.03.49
10
63
62
42
1
178
3.53.69
0
15
21
60
14
110
3.74.00
0
3
12
58
72
145
24
117
110
160
90
501
Database Systems Journal, vol. X/2019
5.1 Preprocessing
Often collected data is not understandable,
inconsistent, lacking in important criteria or
can contain various errors. Preprocessing
makes data understandable by various
process and solve those issues. Hence we
have removed inconsistency from our data
set and transformed them into
39
the proper format to use them for analysis.
Also selected important features to reduce
the complexity of our process. For this we
have used a filter method which is feature
selection by gain ratio. Table-3 shows the
comparison selected features-
Table 5. List of selected features. Here we can see for CGPA programming knowledge is an
important factor. For programming skill self-study, patience and skill of higher level
mathematics is most important.
Selected Features (For
Gain Ratio
Selected Features (For
Gain Ratio
CGPA)
CGPA)
ProRatings
0.474
SSR
0.279
PaRatings
0.465
PaRatings
0.268
GHSC
0.457
HMATH
0.255
SSR
0.452
Cpro
0.246
Cpro
0.448
MSOlym
0.224
GSSC
0.439
NOC
0.212
HMATH
0.439
CGPA
0.211
MSOlym
0.410
BMath
0.183
NOC
0.407
ICTResult
0.182
IBReason
0.404
IBReason
0.172
ICTResult
0.389
GHSC
0.168
uses absolute values and gives intuition (the
5.2 Data Mining and Model Generation
“average error”).
Using Various Algorithms
Root Mean Square Error (RMSE): refers to
For predicting the CGPA and programming
the standard deviation of the prediction
skill we have used several algorithms i.e.
errors. By using RMSE we can tell how
SVR, C4.5 etc. While predicting two
concentrated the data is around the line of
standard evaluation metrics (MAE, RMSE)
best fit.
are used as cost function to evaluate the o
Now we will see the results of regression
performance of our prediction.
algorithms for predicting the CGPA and
Mean Absolute Error (MAE): measures the
programming skill in respect to these cost
difference between two continuous values. It
functions.
Table 6. Error evaluation for Regression (CGPA prediction)
Evaluation Criteria
Multiple Linear
C4.5
Regression
MAE
0.1180
0.1555
SVR (Linear
kernel)
0.1123
SVR (Poly
kernel)
0.1409
RMSE
0.1604
0.2117
0.1712
0.2517
Table 7. Error Evaluation for Regression (Programming Skill Prediction)
Evaluation Criteria Multiple Linear
C4.5
SVR (Linear
Regression
kernel)
MAE
0.5226
0.4242
0.5036
SVR (Poly
kernel)
0.4596
RMSE
0.6584
0.6899
0.7881
0.6880
40
Factors that contribute programming skill and CGPA as a CS graduate: Mining Educational Data
Taking all these results to account we can
say MLR is best for predicting CGPA and
SVR with linear kernel is bes…
Purchase answer to see full
attachment