ISSN (Online) 2581-9429

IJARSCT
International Journal of Advanced Research in Science, Communication and Technology (IJARSCT)

Volume 3, Issue 2, March 2023

Impact Factor: 7.301

Personal Career Recommendation System

Dr. A Syed Mustafa1, Prof. Anusha KV2, Deepak N3, Goutham C4, Karthika Hiremath5, Lokesh SB6
Professor and Head of Department, Department of Information Science and Engineering 1
Assistant Professor, Department of Information Science and Engineering2
Students, Department of Information Science and Engineering3,4,5,6
HKBK College of Engineering, Bangalore, Karnataka, India

Abstract: As students are going through their academics and pursuing their interested courses, it is very
important for them to assess their capabilities and identify their interests so that they will get to know in
which career area their interests and capabilities are going to put them. This will help them in improving
their performance and motivating their interests so that they will be directed towards their targeted career
and get settled in that. Also, recruiters while recruiting candidates after assessing them in all different
aspects, these kinds of career recommender systems help them in deciding which job role the candidate
should be kept in based on his/her performance and other evaluations. This paper mainly concentrates on
the career area prediction of computer science domain candidates.

Keywords: Carrer, SVM, XGBoost, Machine Learning

I. INTRODUCTION
Nowadays, students are often facing a dilemma in deciding to choose a career in their life. There are several factors that
influenced students when choosing their career path as their Personal aptitudes, educational achievement, and their
environment. Upon completing their first degree or undergraduate students at the university, students are normally
starting to consider a career path that may suit their skill and potential the best. However, many students made the
wrong decision in selecting their career due to the lack of experience, help, and advice from friends and relatives,
parents and lecturers, or career counseling. It is necessary to deliver career guidance in several ways like courses,
training, and seminars that offer group experiences in future career planning and group or individual counseling
activities. The factor that may cause students not to be successful in their careers is the wrong choice of a job that suits
them. It requires a decision-making process at an early stage. So this system gives a recommendation for students about
their careers based on their academic result and their abilities.

II. ALGORITHMS
2.1 Support Vector Machine
Support Vector Machine is an acronym. It is a supervised machine-learning approach that is typically applied to both
classification and regression-type issues. Various classification issues are where this is primarily applied. The standard
algorithmic step is to plot each data point in an n-dimensional space, where n is the number of features and a feature's
value is the value of a specific coordinate. The following step is to classify by obtaining the hyper-plane that sharply
divides the two classes. Practically speaking, kernels are used to implement SVM algorithms. There are three different
types of SVM, and the linear SVM hyperplane is calculated or discovered by using linear algebra to the problem. The
realization is that SVM can be rephrased by using the inner product of observations.

2.2 XG BOOST
Extreme Gradient Boosting is referred to as XGBoost. The implementation of gradient boosting algorithms is called
XGBoost. It is accessible in a variety of tools mats, including tools, a library, etc. It primarily concentrates on
computational efficiency and model performance. It significantly cuts down on the amount of time and significantly
improves the model. Its implementation includes recent additions like regularisation in addition to capabilities seen in
sci-kit-learn and R versions. Gradient boosting using both L1 and L2 type regularizations is referred to as regularised
gradient boosting. The following are the key benefits that the algorithm's implementation offers: Automatic handling of

Copyright to IJARSCT DOI: 10.48175/IJARSCT-8675 135

www.ijarsct.co.in
 ISSN (Online) 2581-9429
IJARSCT
International Journal of Advanced Research in Science, Communication and Technology (IJARSCT)

Volume 3, Issue 2, March 2023

Impact Factor: 7.301
missing values with sparse aware implementation, block structure to facilitate parallel tree construction, and ongoing
training to support further enhancing a model that has already been fitted on the new data.

2.3 Decision Tree

A popular and straight-for e-learning categorization challenge is the use of decision trees. For several sophisticated
algorithms, like bagging, gradient boosting, and random forest, decision trees provided the fundamental building
blocks. The improved form of the previously stated XG Boost method this broad decision tree CART, C4.5, C5, and
ID3 are the three most popular decision trees. In the event that the variable is numerical, a node represents a split on the
input variable (X). An output variable (y) that is essential for prediction is present in the leaf, also known as the tree's
terminal nodes. Selecting a root node is the first step in the usual decision tree process. Before the split, figure out each
node's information gain or entropy. All paragraphs must be indented. All paragraphs must be justified, i.e. both left-
justified and right-justified.

III. IMPLEMENTATION
3.1 Data Collection
One of the biggest and most crucial jobs of every machine learning project is data collection. as a result the input data is
fed to the algorithms. Therefore, the accuracy and efficiency of the algorithms depend on how well the data is collected
and how accurate it is. The result will be the same as the data. Numerous factors, including academic performance in
multiple disciplines, specialties, programming, analytical skills, memory, relationship status, interests in sports,
contests, hackathons, workshops, and books, among others, are needed to predict a student's career. All of these
characteristics are taken into account since they are crucial in determining how far a student will advance in a certain
vocational field. Many methods exist for gathering data. Some information is gathered from employees of various
organizations, LinkedIn, and Colleges

3.2 Data Pre-Processing

Making the data meaningful is a crucial effort that goes hand in hand with gathering the data. There may be many null
values, invalid data values, and undesirable data in the data obtained through various methods because it will be in a
disorganized manner. removing all of this data and substituting them with more accurate or suitable data. The
fundamental processes in preprocessing data include identifying null and missing data, eliminating it, and replacing it
with some predefined alternate values. Even acquired data could have entirely useless values. It might not be in the
precise shape or manner in which it was intended. To make the meaning of the data intelligible and helpful for future
processing, all such cases must be validated and replaced with alternative values. The storage of data must be
organized.

3.3 One Hot Encoding

OneHot Encoding is a method for providing categorical values found in the collected data to machines by converting
them to numerical or other ordinal formats. Improving prediction outcomes via learning algorithms. Categorical values
are transformed using the OneHot encoding technique into a format that is most suitable for feeding into different
machine learning algorithms. Nearly any machine learning algorithm is compatible with this algorithm. Only the a few
algorithms, like random forest, effectively handle categorical values. In these circumstances, one hot encoding is not
necessary. Although the OneHot encoding process may appear challenging, most contemporary machine learning
algorithms handle that. This article clearly explains the procedure: In a data set, for example, if there are yes and no
values, integers respectively 1 and 0.

Copyright to IJARSCT DOI: 10.48175/IJARSCT-8675 136

www.ijarsct.co.in
 ISSN (Online) 2581-9429
IJARSCT
International Journal of Advanced Research in Science, Communication and Technology (IJARSCT)

Volume 3, Issue 2, March 2023

Impact Factor: 7.301
IV. ARCHITECTURE DIAGRAM

V. LITERATURE SURVEY
5.1 Personal Factor
Personal factors including student behavior, such as feelings, activities, as well as the finances of the pupils, are
important components of understudy academic performance. age and sexual orientation, which impact how they
perform. The most often employed parameters for prediction are age and gender since they are considered internal
causes of variability and are simple to define and assess. The analysts looked at how frequently psychometric factors
might affect how well students performed.

5.2 Academic Factors

Academic factors are descriptors used to describe variables that explain a student's performance on the academic track
at a university. Given its enormous influence on the future of training, the cumulative grade point average (CGPA) is
the primary factor that has been used most frequently student author took into account the student GPA. The most
significant factors in forecasting a student's future academic performance, according to the authors, are their past
academic performance and their parents' educational backgrounds. Others investigated how past academic success
affected how well pupils will do going forward.

5.3 Financial Factors

Financial aspects Related elements imply that the parents' financial capacity to support their children's education and
guide their future careers. A few specialists looked at the relationship between educational programming, parental
instruction, and salary.

5.4 Family Factors

Family variables have an impact on an individual's educational background, ability to help their children with their
education, and ability to create an environment that is conducive to learning. Results showed that regardless of whether
students’ school has an impact on students’ performance, parents' efforts have a crucial role in predicting grades. The
Copyright to IJARSCT DOI: 10.48175/IJARSCT-8675 137
www.ijarsct.co.in
 ISSN (Online) 2581-9429
IJARSCT
International Journal of Advanced Research in Science, Communication and Technology (IJARSCT)

Volume 3, Issue 2, March 2023

Impact Factor: 7.301
most important factors in predicting an understudy's future scholastic presentation in the art hour were determined to be
their former academic performance and their parents' teaching background. Studies often examine how parents have an
impact. education background, income, and the impact of students' interpretation on academic success.

5.5 Institution Factor

The factors that affect this classification have to do with the academic program and resources that the foundation
provides for the highest academic performance of its trainees. The authors looked at how psychometric factors
frequently affect how candidates present themselves.
JOURNAL NAME AUTHOR TITLE& METHOD REMARKS
& YEAR
Hindawi.2022. Qingwan, Linye Career Recommendations for By improving the Activation
students based on Deep function, pooling strategy, and
Learning and Machine loss function in the algorithm,
Learning. A hybrid CNN model the quality of model prediction
is proposed for employment is greatly improved.
recommendation of college
students.
International Journal Sushma Koushik N, Educational Career The recommendation system is
of Advanced Chandana M S Recommendation System built in Python because it is
Research in Lavanya V, Suhas Y, Using Machine Learning. easy and efficient to put into
Computer and HarshithaV The problems of cold start, effect algorithms on exclusive
Communication trust, and privacy are solved in operating systems.
Engineering. 2021. this approach
Multidisciplinary MinNie, Career Choice Prediction Based A noval regularization
Digital Publishing Zhaohui Xiong on Campus Big Data—Mining Item was introduced by
Institute. Ruiyang Zhong, Wei the Potential Behavior of To bridge the gap between the
2020. Deng and Guowu College Students Proposeda real-world examples and
Yang Prototypical cluster center prototypical cluster centers.
generation approach to use the
priori information from each
college
Research gate.2017 Bharat Patel, Varun A career path Recommendation The system can run on cloud
Kakuste, Magdalini on framework A novel career infrastructure using apache
Eirinaki path framework for spark or Hadoop cluster which
personalized job and skills can handle more number of jobs
Recommend youngs, focusing description and more users. To
on students and young increase the system capability at
professionals is presented. any time, new machines can be
added to the existing cluster
without affecting the running
application.
International Journal Dr. Sadasivam R, Student carrer prediction RF contains various choice
of Health Science Paramasaivam S, Student suggestion was threes different subsets of the
Journal of Health Prakash raj N, proposed to help student career given datasets and taken the
Sciences. 2022. Saravanam M prediction in concluding there normal to work on the precision
abilities in which they are solid of that datasets. It tends to be
and frail. utilized for both Arrangements
and Regression issue in ML

Copyright to IJARSCT DOI: 10.48175/IJARSCT-8675 138

www.ijarsct.co.in
 ISSN (Online) 2581-9429
IJARSCT
International Journal of Advanced Research in Science, Communication and Technology (IJARSCT)

Volume 3, Issue 2, March 2023

Impact Factor: 7.301
VI. METHODOLOGY

The system was created using Python. The system produces after receiving input from the datasets. As a result The
steps that make up the system-building process are done in order.
Obtaining the dataset
Cleaning the Dataset
Selecting for feature dataset
Build a model
Predict outcomes using the model

VII. RESULT
All three algorithms are tested and trained on the data, and SVM topped all others in terms of accuracy, scoring 90.3
percent, followed by XG Boost at 88.33 percent. As SVM All subsequent data predictions are picked to be followed by
SVM because it provided the highest accuracy. Thus, a web application is created to provide the student's input
parameters and the end A prediction is produced and shown. SVM is indeed the background method in use, and fresh
predictions are continuously added to the dataset to increase accuracy.

Copyright to IJARSCT DOI: 10.48175/IJARSCT-8675 139

www.ijarsct.co.in
 ISSN (Online) 2581-9429
IJARSCT
International Journal of Advanced Research in Science, Communication and Technology (IJARSCT)

Volume 3, Issue 2, March 2023

Impact Factor: 7.301
VIII. CONCLUSION
We looked at how college students choose their careers based on their professional expertise, behavior consistency, and
other associated behaviors The study has also provided a number of crucial insights for enhancing the model.
We have suggested a prototype approach for creating cluster centers that takes advantage of the past knowledge from
each college. In order to bridge the gap between real-world instances and a prototypical number of clusters, we
introduced a unique normalization item, which is motivated by the cluster assumption that samples in the same cluster
should have the same label. Multiple experiments' findings show that our method is superior to other methods for
predicting professional choice.

REFERENCES
[1]. https://towardsdatascience.com/building-a-job-recommender-for-non-technical- business-roles-via-nlp-and-
machine-learning-626c4039931e
[2]. https://www.researchgate.net/publication/358791179_Career_Recommendation_for_College_Students_Based
_on_Deep_Learning_and_Machine_Learning
[3]. https://www.researchgate.net/publication/269272063_Career_path_recommendation_system_for_UiTM_Perli
s_students_using_fuzzy_logic
[4]. https://ieeexplore.ieee.org/document/6869553
[5]. https://www.semanticscholar.org/paper/Student-Career-Prediction-Using-Decision-Tree-and-VidyaShreeram-
Muthukumaravel/28490efaca5b4cd1c215fa79dec03f14f58c17c4
[6]. https://ieeexplore.ieee.org/document/9137992 https://www.ijraset.com/research-paper/career-
recommendation-using-ann
[7]. P.KaviPriya, “A Review on Predicting Students’ Academic Performance Earlier, Using Data Mining
Techniques”, International Journal of Advanced Research in Computer Science and Software Engineering.
[8]. Ali Daud, Naif Radi Aljohani, “Predicting Student Performance using Advanced Learning Analytics”, 2017
International World Wide Web Conference Committee (IW3C2).
[9]. Nikita Gorad ,Ishani Zalte, “Career Counselling Using Data Mining”, International Journal of Innovative
Research in Computer and Communication Engineering.

Copyright to IJARSCT DOI: 10.48175/IJARSCT-8675 140

www.ijarsct.co.in