University Recommendation Support System using ML Algorithms

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 01 | Jan 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 452
University Recommendation Support System using ML Algorithms
Dipti Babel1, Ashutosh Rathi2, Sharvari Rodge3, Saurabh Thorat4, Prof.M.D.Salunke5
1-5Department of Computer Engineering, JSPMNTC (RSSOE), Pune, India, Savitribai Phule Pune University
-------------------------------------------------------------------------***----------------------------------------------------------------------
Abstract: For a prospective undergrad student, choosing
which universities to apply to is a mystery. Students often
wonder if their profile is good enough for a particular
university. In this article, we have addressed this problem
by modeling a recommendation system based on various
classification algorithms. The required data was extracted
from www.edulix.com and a dataset was created with
profiles of students with admission/rejection from 45
different universities in the US.Based on this data set,
various models were trained and a top 10 university list
was proposed in order to maximize the chance of a student
being accepted into this university list.
1.INTRODUCTION
Number of students completing postgraduate studies
abroad. The process for obtaining fully funded graduate
study opportunities is very systematic and competitive.
Many students apply to different universities in different
countries. The universities offer admission to suitable
candidates based on their academic profile, Test results,
work experience and research.
But in this entire process, college selection is the most
important step in applying for admission to college.
Unique platform that can shortlist the universities/
colleges that attract applicants. The findings from the
database of selected applicants are sufficient to find
answers to questions such as: Which factors determine the
funding opportunities for applicants at a particular
graduate school? Student categories are usually assigned
to the M.Sc. o Ph.D. from graduate school? What are the
key factors required to receive graduate funding after
selecting the right graduate school? Data mining
techniques are very helpful in uncovering this kind of
hidden knowledge about basic and complex data types.
The main objective of this research is to design and
develop a referral system for college entrance applicants
that can help them select the graduate school that fits their
full profile based on academic data from students who
have already had the opportunity to apply abroad to study.
The proposed system analyzes the data from these
datasets, selects the data characteristics, executes the
classification algorithm of the Support Vector Machine
(SVM) and the machine learning algorithm K Nearest
Neighbours (KNN) on it and suggests suitable universities
to the applicants accordingly.
2. Need of Recommendation Systems
In today's fast-paced world, any technological innovation
affects the importance of higher education, especially
those that act as hubs for the latest research and trends.
Since then, America has become one of the top travel
destinations for any student around the world. If you want
to undertake postgraduate studies in the United States of
America, choosing a suitable university and admission are
a challenge. no actual statistical relationships. From the
student's point of view, the costs for the application and
the commitment to the process are also high. In order to
guide students efficiently, the college referral system was
developed based on student contributions. Since the
problem is extensive, a selected list of 45 universities has
been considered for the sake of simplicity.

3. Literature Review
Much work has been done in the past on the use of data
mining techniques in education. Few recommendation
systems have been developed to suggest courses and
colleges based on the student's academic performance.
These systems used the decision tree classifier and fuzzy
media clustering techniques using the WEKA tool. and it
should help students choose a sequence that suits their
skills. Another referral system has been developed to
assist students in their academic path.in making decisions
about course choice based on the student's schedule,
order, and teachers. The model was trained on the
Basis of data from the last 7 years for a specific
university and the classifiers for each subject were
modeled on the basis of the cumulative GPA, some referral
systems have been modeled to help the university get to
know their students by tracking their time, extracurricular
activities and accomplishments, and academic potential,
and helping them identify and categorize students as
needed using two algorithms and Kmeans . However, there
was no access to all of the data sets that were used in the
above-mentioned work. Although there are similarities
with the subject covered in this document, it is not
appropriate to directly compare the results with previous
work as the data set used in this document is completely
different.
4. Proposed Architecture
4.1. Data Set Collection
The first step in setting up a recommendation system is
the identification of the data set. For this special problem,
the academic details and background information
provided in the application process represent the basic
data. Classification model for the recommendation system,
these data must be sorted with appropriate labels. These
basic data for the application process are not readily
available on the Internet for direct consumption. Although
there were few forums that gave important information
related to the scores, the most distinctive information was
about student research. Interest and knowledge on a
particular topic are unknown. However, this entire
approach is based on making the most of the information
available. Figure 2 shows the different number of
admissions for each graduate college depending on the
Bachelor universities. The University of Mumbai (1587),
the National Institute of Technology (1467), Visvesvaraya
Technological University (1426) and Anna University
(1032) were found to be among the undergraduate
universities with the most admissions Universities The
'Edulix' forum is one of the most popular forums for
students aiming for postgraduate studies. This is the
contact point for students who would like to take part in
discussions and inquiries about all information about
graduate studies. This forum essentially gathers the
academic details of its users in order to compare their
profile with previous experiences. From all of this data,
some data such as the candidate's undergraduate
university, CGPA, GRE and number of research
publications, work experience, etc.
4.2. Data Scraping
Initially the list of forty five universities was narrowed
down, and had enough information to be scrapped.
Universities with skew data were born down. Then a
crawler was engineered to induce the list of scholars and
therefore the links to their profiles on Edulix. Once the
distinctive set of students was identified, the data was
scraped from every profile then the desired data was
extracted from the hypertext markup language by
mistreatment the python library ’Beautiful Soup’. The
tabular structure of Edulix’s net page, helped to spot the
required data labels and points. The standard way of

accessing the desired parts by mistreatment the XPath
failed to estimate for this case, as a result of the hypertext
markup language was distorted in several cases.
4.3. Data Pre-processing
About 45000 samples of information were obtained as a
result of scraping. Every sample corresponds to the profile
of a student. The information points extracted enclosed
GPA, college Boy University, GRE verbal score, GRE
quantitative score, GRE analytical writing score, variety of
journal publications, number of conference publications,
trade expertise, analysis experience, situation experience
and the following major. Cleansing the data of
undergraduate universities had to be done, since this field
was simply a text box and not a get field. Thus, input from
totally different students created anomalies and this was
corrected by trimming the string and removing areas
found in them. The GRE scores (Verbal, Quantitative and
AWA) were additionally clean since they contained
immeasurable each previous and new versions of the
examination. Equally the purpose
average|GPA|standard|criterion|measure|touchstone}
scores out there were supported totally different point
systems, so all the GPA scores were uniformly scaled to
four point scale. Also, bound categorical options just like
the student’s college boy university and department to
which they apply were thought of as separate features. a
complete of 1435 distinct undergraduate universities and
fifty-three distinct majors were obtained when filtering
and every of those were used as binary options.
4.4. Feature Extraction
The most necessary property of a feature is its correlation
with the anticipated output. alpha analysis was done by
plotting the feature values for 2 totally different
universities and observing their variation. Variation of
options CGPA and GRE for two different
universities(Purdue and NJIT),has been shown in Figure
three and Figure four respectively.
Initially, once all the features within the information set
were thought-about the accuracy was relatively low(40%).
The forward choice algorithm was accustomed to choose
the simplest set of features for the model. Within the 1st
iteration of the algorithm, the single best feature was
known that best describes the variance within the data.
Within the second iteration, the simplest feature was
mounted and also the next best feature was found. This
method was recurrent until the accuracy now not
improved. Supported this method, collegian university,
analysis experience, GRE, and grade point average were
found to be the foremost effective options. once
mistreatment the forward choice algorithm, the accuracy
improved. Throughout this process, a scenario arose, when
the accuracy didn't show any improvement, even if the
best features were chosen. This was as a result of the
numerical features like CGPA Associate in Nursing GRE
scores were supported totally different scales, and then
had an adverse implication on the model. However, once
scaled from zero to one, there was a major improvement
within the accuracy. Hence, all the numerical variables
were then normalized to a scale of 0 to 1 by mistreatment
the subsequent formula,
X = X − Xmin
-----------------------------
Xmax − Xmin
wherever X is the price of any feature.
4.5. KNN
The K Nearest Neighbours algorithm is a nonparametric
method of classification and regression. In the case of the
KNN classification, the output is a class membership. An
object is classified by the majority of its neighbors, and the
object is assigned to the most common class among its k
nearest neighbors (k is a positive integer, typically small) If
k = 1 then the object is simply assigned to class des closest
unique item Neighbor.

4.6. SVM
Support Vector Machine(SVM) could be a supervised
machine learning rule used for each classification ANd
regression. tho' we are saying regression issues
additionally its best fitted to classification. The target of
the SVM algorithm is to search out a hyperplane in an N-
dimensional area that clearly classifies the information
points. The dimension of the hyperplane depends upon the
quantity of options. If the number of input features is two,
then the hyperplane is simply a line. If the number of input
features is three, then the hyperplane becomes a 2-D
plane. It becomes troublesome to imagine once the
number of features exceeds three.
5. CONCLUSIONS
The best university may be suggested to the scholars as
per their requirement. Will be Associate in Nursing
intelligent recommendation system that helps students to
see their eligibility for a University supported University
admission criteria. It’ll conjointly embrace several
parameters like GRE, TOFEL score, University rank,
budget, weather, and so on for recommending the right
university.
A substantial variety of scholars are conferred with the
chance to pursue instruction once the completion of their
undergrad studies in countries completely different than
their home countries. The records of those students that
have with success gained admission can be constructive
and worthy for alternative students hoping to achieve
admission and facilitate them in their call making. Data
processing and Machine Learning are the paradigms that
may explore and supply exemplary results. Therefore the
past records of fortunate graduate candidates hold utmost
importance in choosing acceptable instruction institutes
for graduate applicants who want to pursue higher studies.
Conclusive of this research, we've planned a graduate
university recommendation system that may apply SVM to
classify a graduate university that's probably appealing to
an applicant and also the KNN algorithm program to come
up with a university with similar necessities and
qualifications.
ACKNOWLEDGEMENT
We would like to thank Prof. Salunke M.D for his guidance
and suggestions.
REFERENCES
[1] A Sabic, Md El-Zayat
https://ieeexplore.ieee.org/document/8025933 2nd IEEE
International conference in April, 2010
[2] M Hassan, Shibbir Ahmed, Deen Md.Abdullah
https://ieeexplore.ieee.org/document/7434430 5th IEEE
International conference, 2016
[3] Murtala Ismail, Usman Haruna, Garba Aliyu, Idris
Abdulmumin
https://ieeexplore.ieee.org/document/7760053 IEEE
[4] Huma Samin, Tayyaba Azim
https://ieeexplore.ieee.org/document/8693719 IEEE
Access Volume: 7, 2019
[5] Can Ozturan, Suleyman Uslu, Mehmet Faith Uslu
https://ieeexplore.ieee.org/document/7991812 10th IEEE

University Recommendation Support System using ML Algorithms

More Related Content

Similar to University Recommendation Support System using ML Algorithms (20)

More from IRJET Journal (20)

Recently uploaded (20)

University Recommendation Support System using ML Algorithms