Student’s Career Interest Prediction using Machine Learning

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 11 | Nov 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 533
Student’s Career Interest Prediction using Machine Learning
1Prof. Priyanka Shahane, 2Prutha Rinke, 3Taniksha Datar, 4Soham Badjate
1Assistant Professor, SCTR’s Pune Institue of Computer Technology,411043
2,3,4Student, SCTR’s Pune Institue of Computer Technology, 411043
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Many students are known to encounter
uncertainty about their career choices and the
marketability of the interests they own and wish to pursue.
Due to the availability of many domains in the evolving
world, the complexity of this issue is increasing. So, in this
competitive world with so many domains and skills to
work on, it is nearly impossible for a human to explore all
the domains and get knowledge of the skills in such a short
period so as to acquire an idea about what domains and
skills would interest the individual, which can help in
building the overall career path. It can be predicted based
on an individual's academic and non-academic history and
also keeping in mind the individual's interests, school
performances, and perspective on future careers. It is a
critical data set for research, as people tend to explore
many areas along the way. This research aims at studying
various machine learning based techniques for students’
career interest prediction.
Key Words: (SVM, OneHot Encoding, Decision Tree,
XG Boost, K-Nearest Neighbor, Educational Data
Mining, Regression Models, Supervised Learning,
Unsupervised Learning)
1.INTRODUCTION
In today’s competitive world it is very difficult for
students to understand their interests and find their
suitable career. There are numerous fields to choose
from. Choosing from this huge plethora of career options
is a real challenge before the individuals today. To
compete and reach the goals of the students it is very
important for them to plan and organize from the initial
stages of their lives. For this it is necessary to constantly
evaluate their performance, identify interests and keep
track of how close they are to their goals and also assess
if they are on the right track towards their target. This
helps them to improve themselves and also pre evaluate
their capabilities before going to the career peak point.
Recruiters while recruiting people into their
companies evaluate candidates on different parameters
and then draw a final conclusion to select an employee
or not. If selected they have to find the best suited role
and career area for the selected employees. Various
career recommendation systems, job role
recommendation and prediction systems are being used
in different private performance evaluation portals like
Co-Cubes, AMCAT. These portals evaluate the students
technically and suggest the students and companies
job roles best suited for them based on their performance.
Various factors including abilities of students in sports,
academics, extracurriculars and also their hobbies,
interests, skills are also taken into consideration.
We have used advanced machine learning algorithms
like SVM, Random Forest decision tree, XG boost for
predictions. After training and testing the data with these
algorithms we take the most accurate results into
consideration for further processing. This paper deals
with various advanced machine learning algorithms that
involve classification and predictions, which are used to
improve the accuracy for better prediction and also
analyzing these algorithm’s performances.
1.1 Problem Definition and Scope
A problem that never gets resolved in a student's life
are their career choices. With various career paths as
options and choosing out of these is a difficult task for
them. It is impossible for them to explore each and every
domain present in the word and choose the most suitable
of these. And thus, a model is developed based on the
student's performance and interests and is implemented
to find out the most suitable career option for them.
This consists of excellent potential as it is seen as a
benefit to the younger generation for their career choices
and preferences. An accurate prediction model can help
not just young age but also retired adults to work on the
preferable career choice suggested to them and contribute
to society for their social well-being. This model would
help people to dig out their choices from a valley of career
paths and provide them with suitable jobs which would
help them support themselves and also increase the
global economy.
2. LITERATURE SURVEY
[1] Personalized Career Path Recommender System
for Engineering Students Opting for a university
specialization is a grueling decision for all academy
scholars. Due to the lack of guidance and limited online
coffers, scholars' opinions depend on the private
comprehension of family and friends. This increases the
threat of high university dropout rates, and students
changing their university disciplines and choosing an
irrelevant career path.

To address the downsides mentioned, this exploration
paper presents a Personalized Career- path
Recommender System( PCRS) to give guidance, and the
right choice, and help all academy scholars choose their
separate disciplines. The main idea of PCRS is to mimic
the part of professional counsels who help scholars make
this hard decision by assaying their academic and
particular interests. The design of PCRS is grounded on a
fuzzy-sense intelligence with two main input
parameters; academic performance and specific profile.
[2] Prediction of student's academic performance
using machine learning algorithms Learning analytics
and supportive learning are known to be emerging
research areas in today’s era of big data, data mining,
machine learning, and artificial intelligence to facilitate
students’ learning. Student education is crucial to the
sustainable development of society as students learn
knowledge from schools and through extracurricular
activities and create abilities to contribute to the
community. There are many students who have
progressed to higher levels of education and earned
degrees like Ph.D. while many graduate every year.
However, there are some students who marginally pass
the course and some who fail the course as well who are
required to have a compulsory retake the same course.
This paper has been proposed as an improvised
conditional network-based deep support vector machine
(ICGAN- DSVM) algorithm.ICGAN focuses on addressing
the issue of low data volume that is using fewer datasets
by mimicking new training datasets whereas DSVM
extends SVM from shallow learning to deep learning.
DSVM takes the advantage of a small dataset, as a key
difference in comparing with the traditional deep neural
network which makes it more efficient to work and
execute.
[3] A Machine Learning Based Approach for
Recommending Courses at Graduate Level Students
often face confusion during their academic career
regarding the choice of the courses they make and their
future scope, hence they need proper guidance for the
same which not everyone is accessible. This paper is
used to propose a system that can be used for
recommending courses to students according to their
benefits and interests. The paper focuses on a higher-
level or graduate level of study. It utilizes techniques of
Data Mining and Machine Learning in order to give
accurate results to the students. The factors taken into
account are based on the student's performance and
their preferred interest and skills. Machine learning
methods like neural networks and various learning
algorithms can help in a student’s career and their best
interest.
[4] Analysis Optimization K-Nearest Neighbor
Algorithm with Certainty Factor in Determining Student
Career A career is a series of development or progress in
their respective professions experienced by every
human being. These careers are manifested as a job title
and job work and something related to the professional
world and academic growth. The crucial element of a
career is the development or progress on every level of
life experienced by humans. K-Nearest Neighbour is a
method for data. Classification and organization while the
certainty factor is an uncertain decision-making method
depicted in this paper. This study uses datasets like
students’ interests, talents and exam scores to predict a
career-appropriate decision for each student. The Student
career prediction system was formed by combining two
methods which are K-Nearest Neighbor and the Certainty
Factor. It was expected that the two-way analysis might
provide a piece of better formation for students in
determining their careers with accurate results. The K-
Nearest Neighbor method received a value derived from
the Certainty Factor which is apparently beneficial in
predicting career prediction. A system was proposed that
uses the KNN-certainty factor method to predict the
student's career with accuracy.
[5] Analyzing the learning of individual and
suggesting field of study using Machine Learning. The
learning style can be defined as the way a student prefers
to acquire, process, and retain the knowledge received
from external sources. The prominent learning style
classification model is called the VAK model. According to
this theory visual, kinesthetic and auditory are the three
major kinds of learning styles. Many kinds of research
have shown that people prefer more than one way of
learning and memorizing, hence categorizing a person as
one of the above types as done in traditional methods is
not reliable. A method that is used to identify learning
styles more accurately is required. Machine learning is
applied to achieve our aim in the most efficient way. Once
we have accurate information about learning styles and
methodologies, we can use it to suggest career options
and predict outcomes. This research mainly aims at
predicting the learning style combinations of students and
suggesting a field of a domain using algorithms like k-
means, SVM, and decision tree.

Sr.
No.
Paper Best classification
technique
Other techniques
tested
Dataset Performance
1 Personalized
Career Path
Recommender System for
Engineering Stu- dents.[1]
PCRS technique Fuzzy-Logic System Personal Information,
Academic Performance
and Personality Type
91%
specificity and 90%
average accuracy
2 Predicting
Students’
Performance using
Generative
Adversarial and Deep
Support
Vector Machine [2]
ICGAN-DSVM
algorithm
Methodology of
ICGAN DSVM
Student Performance,
IGAN and CGAN datasets
Accuracy not mentioned
3 Prediction
of students’ academic
performance using machine
learning algorithms [3]
RF, NN and SVM
algorithms
DM Model, Predictive
Model, Descriptive
model
Student Information
System (SIS) of a State
University, Turkey
Accuracy and
susceptibility
4 Predicting Stu- dents’
Academic Performance at
Secondary and Intermediate
Lev- el Using
Machine Learning [4]
Recurrent neural net-
works
Algorithms of
Machine Learning,
classifiers based on
Deep-learning and
regression models
BISE Peshawar, of SSC &
HSSC were acquired from
6
different Boards
Accuracy and
susceptibility
5 A Machine Learning Based
Approach for
Recommending Courses at
Graduate Level [5]
Xavier initialization ML
algorithms
From 200 students
engineering colleges with
students’ recent semesters’
performances, interests
73%
accuracy
6 Career Recommendation
Systems using Content
based Filtering [6]
Career Recommendation
frameworks
Content based
filtering
algorithm, Natural
Language
Processing
Percentage in OS,
Algorithms and design,
Average percentage in
programming, software
engineering, electronics,
computer architecture,
mathematics.
No accurate accuracy
7 An Intelligent Ca- reer
Guidance System using Ma-
chine Learning [7]
K-Nearest
Neighbor
Confusion Matrix,
SVM, Naïve Bayes
500 unique values with
skillset of candidates using
multitask classification
technique
No accurate
accuracy
8 Analysis Optimization K-
Nearest Neighbor Algorithm
with Certainty Factor in
Determining Student
Career.[8]
K- Nearest Neighbor Certainty Factor
method
Three datasets Computer
Engineering network,
software Engineering,
Multimedia are used
93.83%
9 Career Trajectory Prediction
based on CNN [9]
CNN Model Architecture RNN
model, DCNN
model
Data Castle China game
dataset
Accuracy and
susceptibility
10 Analyzing the learning of
individual and suggesting
field of study using Ma- chine
Learning.[10]
SVM and Decision tree SVM and
Decision tree
Student Survey conducted
in Amrita Vishwa
Vidyapeetham University
92.8%
accuracy
3. COMPARATIVE ANALYSIS

4.1 Decision trees
4. DIFFERENT ALGORITHMS USED
4.2 Support Vector Machine
Tree based learning calculations are considered to be
one of the leading and generally utilized directed
learning strategies. Tree based strategies enable
prescient models with tall precision, steadiness and
ease of elucidation. Not at all like straight models,
they outline nonlinear connections very well. They are
versatile at tackling any kind of issue at hand
(classification or relapse). Strategies like choice trees,
irregular timberland, angle boosting are being
prevalently utilized in all sorts of information science
problems. Hence, for each examiner it’s critical to
memorize these calculations and utilize them for
demonstrating. Choice tree may be a sort of
administered learning calculation (having a pre-
defined target variable) that's for the most part
utilized in classification issues. It works for both
categorical and nonstop input and yield factors. In this
procedure, we part the populace or test into two
or more homogeneous sets (or sub-populations) based
on most critical splitter / differentiator in input
variables
Fig -1: Decision tree
Types of choice tree is based on the type of target
variable we have. These are of two types:
1. Categorical Variable Choice Tree: Choice Tree
which has categorical target variable at that point it called
as categorical variable choice tree. Illustration:- In over
situation of understudy issue, where the target variable
was “Student will play cricket or not” i.e. YES or NO.
2. Continuous Variable Choice Tree: Choice Tree has
persistent target variable at that point it is called as
Continuous Variable Choice Tree .
Support Vector Machine (SVM) may be a directed
machine learning calculation which can be utilized for
both classification or relapse challenges. In any case, it
is generally utilized in classification issues. In this
calculation, we plot each information thing as a point in
n dimensional space (where n is the number of
highlights you have) with the esteem of each highlight
being the esteem of a specific facilitator. At that
point, we perform classification by finding the
hyper-plane that separates the two classes very well.
Support Vectors are essentially the coordinates of
person perception. Back Vector Machine could be a
wilderness which best isolates the two classes
(hyper-plane/ line). In SVM, it is easy to have a straight
hyper-plane between these two classes. But, another
burning address which emerges is, ought to we ought
to add this include physically to have a hyper-plane. No,
SVM features a procedure called the kernel trick.
These are capacities which takes moo dimensional
input space and change it to the next dimensions.
Fig -2: Support Vector Machine
4.3 One Hot Encoding
OneHot Encoding is a technique by which
categorial values present in the data collected are
converted into numerical or other ordinal format so
that they can be provided to machine learning
algorithms and get better results of prediction.
Simply OneHot encoding transforms categorial
values into a form that best fits as input to feed to
various machine learning algorithms. This
algorithm works fine with almost all machine
learning algorithms. Few algorithms like random
forest handle categorical values very well. In such
cases OneHot encoding isnot required. Process of
OneHot encoding may seem difficult but most
modern day machine learning algorithms take care of
that. The process is easily explained here: For example
in a data if there are values like yes and no., integer
encoder assigns values to them like 1 and 0.This
process can be followed as long as we continue the
fixed values for yes as 1 and no as 0. As long as we

4.4 XG Boost
assign or allocate these fixed numbers to these
particular labels this is called as integer encoding.
But here consistency is very important because if we
invert the encoding later, we should get back the
labels correctly from those integer values especially
in the case of prediction. Next step is creating a vector
for each integer value. Let us suppose this vector is
binary and has a length of 2 for the two possible
integer values. The ‘yes’ label encoded as 1 will
then be represented with vector [1,1] where the
zeroth index is given the value
1. Similarly ‘no’ label encoded as ‘0’ will be
represented like [0,0] which represents the first
index is represented with value
0. For example [pillow, rat, fight, rat] becomes
[0,1,2,1]. This is here imparting an ordinal property to
the variable , i.e. pillow < ret < fight. As this is
ordinally characteristic and is usually not required an
desired and so OneHot encoding is required for correct
representation of distinct elements of a variable. It
makes representation of categorial variables to be more
expressive.
XGBoost is an open-source program library which
gives the angle boosting system for C++, Java,
Python, R, and Julia. It works on Linux, Windows,
and macOS. From the venture depiction, it points to
supply a "Adaptable, Convenient and Disseminated
Slope Boosting (GBM, GBRT, GBDT) Library". Other
than running on a single machine, it too bolsters the
disseminated preparing systems Apache Hadoop,
Apache Start, and Apache Flink. It has picked up
much popularity and consideration as of late
because it was the calculation of choice for
numerous winning groups of a number of machine
learning competitions XGBoost at first begun as a
investigate extend by Tianqi Chen as portion of the
Dispersed (Profound) Machine Learning
Community (DMLC) bunch. At first, it started as a
terminal application which may be designed
employing a libsvm arrangement file. After winning the
Higgs Machine Learning Challenge, it got to be well
known within the ML competition circles.
Before long after, the Python and R bundles were built
and presently it has bundles for numerous other
dialects like Julia, Scala, Java, etc. This brought the
library to more designers and got to be prevalent
among the Kaggle community where it has been
utilized for a expansive number of competitions. It
before long got to be utilized with numerous other
bundles making it easier to utilize within the particular
communities It presently has integrative with scikit -
learn for Python users, conjointly with the caret bundle
for R clients. It can too be coordinates into
Information Stream systems like Apache Start, Apache
Hadoop, and Apache Flink utilizing the dreamy
Rabit and XGBoost4J. The working of XGBoost
denotes eXtreme Gradient Boosting. XGBoost is
implementation of gradient boosting algorithms. It is
available in many forms like tool, library et cetera.
It mainly Page | 31 focuses on model performance and
computational time. It greatly reduces the time and
greatly lifts the performance of the model. Its
implementation has the features of scikitlearn and R
implementations and also have a newly added
features like regularization. Regularized gradient
boosting means gradient boosting with both L1 and L2
type regularizations. The best highlights that the
execution of the calculation gives are: Programmed
dealing with of lost values with meager mindful usage,
and it gives square structure to advance parallel
development of tree and proceeded preparing which
underpins advance boost an as of now fitted
demonstrate on the fresh data. Gradient boosting could
be a method where unused models are made that can
anticipate the errors or remains of past models and after
that included together to create the ultimate
expectation. they utilize angle descendent calculations to
decrease misfortune amid including of unused models.
They back both classification and relapse sort of
challenges. Within the preparing portion by and large an
objective work is characterized. Characterize an
objective work and attempt to optimize it. obj=i=1
Σl(yi,yi^(t))+ i=1 Σ Ω(fi
5. METHODOLOGY
5.1 PROCESS FLOW
Chart -1: System Architecture

5.2 DATA COLLECTION
5.3 DATA PRE-PROCESSING
5.4 APPLICATION OF ALGORITHMS
5.5 TRAINING AND TESTING
6.CONCLUSIONS
Collection of data is the foremost and most
important task of any machine learning project. This
is the data that we tend to feed the algorithms to obtain
the required predictions and results. The algorithm's
efficiency and accuracy depends upon the correctness
and the quality of the data that is feed-ed. For student
career prediction several parameters are taken into
consideration like students educational scores in
varied subjects, specializations, programming and
analytical capabilities,
memory, personal details like hobbies, interests,
sports, competitions, hackathons, workshops,
certifications, books interested and many more. As a lot
of these factors play important role to decide students
progress towards career track, all of these are taken
into thought. Data is collected from different sources.
Some data is collected from employees working in
different organizations, some amount of data is collected
through linkedin api, some amount of data is
randomly generated and some from college alumni
databases.
Collecting the data is a necessary task, but making that
data insightful is also necessary. Data will be collected
from various sources, and there may be a lot of
invalid data values and unwanted data. Cleaning all
these data and replacing them with appropriate or
approximate data and removing null and missing data
are the basic steps in preprocessing data. Even the
collected data may contain complete garbage values.
It may not be in the format or in the way it should be.
All such cases must be verified and replaced with
alternate values to make data meaningful and useful
for further processing. Data must be kept in a specific
format. All of these things are taken into consideration
while data pre-processing.
The next step after data pre-processing is applying the
algorithms to the data and obtaining the results and
observations. After applying the algorithms we need to
analyze and try to improve the accuracy of the
algorithm.
Finally, after data processing and training, the next task
is obviously testing. This is where performance, quality
of data, and required output all appear. Of the massive
data set collected, 80 percent of the data is used for
training and 20 percent of the data is reserved for
testing. Training the machine is the process of
making it learn and giving it the capability to make
further predictions based on the training it took.
Testing means having a predefined data set and
output also previously labeled and the model is tested
whether it is working properly or not. If the maximum
number of predictions are correct, the model will
have a good accuracy percentage and be reliable to
continue using otherwise better to change the model.
Competition in today’s world is heavily propagating day
by day. Especially since it's too common in the current
day’s specialized world. So as to compete and
achieve the ambition of children. It needs to be mapped
and organized from the original stages of their
education. So it's certifiably important to constantly
estimate their performance, identify their interests, and
estimate and mark them to see how close they are to
their goal and whether they are on the correct path for
their target. This assists them with self-improvement
and motivation and makes the right choice. It enables
them to choose a better career path even if their
qualifications are insufficient. It also helps in
pre-evaluating themselves before reaching the career
peak point. Not only that, but recruiters consider
aspirants when hiring new employees and estimate
their diverse parameters and draw a final conclusion to
take an employee or not and if selected, dig up the
best-suited role and career field for them. After
reckoning all the factors, the total number of
parameters that were taken into consideration to
solidify the model.
An important web operation can be developed where
inputs aren't given directly rather student parameters
are taken by assessing them through different
assessments and quizzing. Specialized, analytical,
logical, memory-based, psychometry and common
awareness, interests, and skill-based tests can be
aimed and parameters are composed through them so
that effects will be indisputably on-target and the
system will be more dependable to exploit.
Also, decision trees have many confines like
overfitting, no pruning, and lack of capability to deal
with null and missing values, and many algorithms
have problems with a huge number of values. All
these can be taken into consideration and indeed
more dependable and more accurate algorithms can
be used. Also, the design will be more important to
depend upon and indeed more effective to depend
upon.

REFERENCES
[1] Kwok Tai Chui, Ryan wen liu, Mingbo Zhao,
“Predicting Students’ Performance With School and
Family Tutoring Using Generative Adversarial
Network-Based Deep Support Vector Machine”.
[2] Nunsina, Tulus, Zakarias Situmorang,” Analysis
Optimization K-Nearest Neighbor Algorithm with
Certainty Factor in Determining Student Career”,
2020 3rd International Conference on Mechanical,
Electronics, Computer, and Industrial Technology
(MECnIT).
[3] Vignesh S, Shivani Priyanka C, Shree Manju H,
Mythili, “An Intelligent Career Guidance System
using Machine Learning”, 2021 7th International
Conference on Advanced Computing &
Communication Systems (ICACCS).
[4] Namratha M, Tanya V Yadalam, Vaishnavi M
Gowda, Vanditha Shiva Kumar, Disha Girish,” based
Filtering”, Proceedings of the Fifth International
Career Recommendation Systems using Content
Conference on Communication and Electronics
Systems (ICCES 2020) IEEE Conference Paper.
[5] John Britto , Sagar Prabhu, Abhishek Gawali
and Yogesh Jadhav, “A Machine Learning Based
Approach for Recommending Courses at Graduate
Level”, Second International Conference on Smart
Systems and Inventive Technology (ICSSIT 2019)
IEEE Xplore Part Number:CFP19P17-ART;ISBN:978-
1-7281-2119-2.
[6] Manar Qamhieh, (Member, Ieee), Haya
Sammaneh, And Mona Nabil Demaidi, (Senior
Member, IEEE), “Personalized Career-Path
Recommender System for Engineering Students”, M.
Qamhieh et al.: PCRS for Engineering Students.
[7] Mustafa Yağcı, “Educational data mining:
prediction of students’ academic performance using
machine learning algorithms”, Yağcı Smart
LearningEnvironments(2022),https://doi.org/10.11
86/s40561-022-00192-z.
[8] Shah Hussain1, Muhammad Qasim Khan,
“Student-Performulator: Predicting Students’
Academic Performance at Secondary and
Intermediate Level Using Machine Learning”,
Received: 22 June 2020 / Revised: 12 April 2021 /
Accepted: 21 April 2021 © The Author(s), under
exclusive license to Springer-Verlag GmbH Germany,
part of Springer Nature 2021.
[9] Miao He, Dayong Shen, Yuanyuan Zhu, Renjie He,
Tao Wang, Zhongshan Zhang, “Career Trajectory
Prediction based on CNN”, Authorized licensed use
limited to : University of Glasgow. Downloaded on
June 02,2020 at 06:05:11 UTC from IEEE Xplore.
[10] Ananthu S Kuttattu, Gokul G S, Hari Prasad, Jijith
Murali, Lekshmi S Nair,” Analysing the learning style
of an individual and suggesting field of study using
Machine learning techniques”, doi
10.1109/ICCES45898.2019.9002051.
[11] Priyanka Shahane, Deipali Gore “A Survey on
Classification Techniques to Determine Fake vs. Real
Identities on Social Media Platforms,” IJRDT, 2018.
[12] Priyanka Shahane, Sneha Kasbe, Rohini
Kasar,M.P. Navle “Ontology Based Information
Retrieval System Using Multiple Queries For
Academic Library,” IRJET, 2018
& Analysis using Machine Learning," 2022
International Conference on Emerging Smart
Computing and Informatics (ESCI), 2022, pp. 1-5,
doi: 10.1109/ESCI53509.2022.9758214.
[13] P. Shahane, "Campus Placements Prediction
[14] P. Shahane, “Detection of Fake Profiles on
Twitter using Random Forest & Deep
Convolutional Neural Network,” International
Journal of Management, Technology And
Engineering, Volume IX, Issue VI, JUNE/2019.

Student’s Career Interest Prediction using Machine Learning

In this document

More Related Content

What's hot

Similar to Student’s Career Interest Prediction using Machine Learning

More from IRJET Journal

Recently uploaded

Student’s Career Interest Prediction using Machine Learning