International Journal of Science and Research (IJSR)

ISSN: 2319-7064
SJIF (2022): 7.942

Smart Career Guidance System using Machine

Learning
Sanil Gokarn1, Sanskar Taware2, Rujuta Vartak3
1Department of Artificial Intelligence & Data Science Vidyavardhini’s College of Engineering and Technology Mumbai, India
Email: sanil.s211029101[at]vcet.edu.in
2Department of Artificial Intelligence & Data Science Vidyavardhini’s College of Engineering and Technology Mumbai, India
Email: sanskar.s211079102[at]vcet.edu.in
3Department of Artificial Intelligence & Data Science Vidyavardhini’s College of Engineering and Technology, Mumbai, India
Email: rujuta.vartak[at]vcet..com

Abstract: Numerous post-secondary graduates encounter a substantial quandary when determining their future trajectories. They
confront the need for guidance in delineating a fitting vocational journey, amalgamating mental, psychological, and academic facets,
along with other pivotal dimensions of their lives. Frequently, they solicit counsel from external sources, precipitating potential
confusion or misguidance. To mitigate this predicament, our System was conceived, empowering students to autonomously navigate
informed decisions sans third-partyinfluence. Functioning as a pivotal platform, the System enables students to delve into their interests
and unearth potential career trajectories congruent with their inclinations. By adeptly collating and scrutinizing pertinent data, the
System employs machine learning methodologies to prognosticate students’ proficiencies and deficiencies, facilitating the revelation of
their domains of mastery.

Keywords: Decision Tree, Support Vector Machine, K NearestNeighbors

1. Introduction 2. Literature Review

Career counseling helps individuals reflect on their passions, As described in [14], a career counseling system has
objectives, competencies, qualifications, and more. It been developed to anticipate the most suitable
aids in linking educational pathways and professional department for an individual by considering their chosen
markets to personal self-awareness. Essentially, career interests and performance in their top two 12th-grade
counseling educates us on how to structure and decide subjects. The system employs K Nearest Neighbor and
on our educational and professional endeavors. It offers Support Vector Machine algorithms, attaining classification
support by streamlining, arranging, and making information accuracies of 94% and 89%, respectively.
about job markets and educational opportunities more
accessible and convenient for us. In contemporary forms, A web-based application for career counseling has been
career counseling integrates insights from diverse academic conceptualized and developed, offering tailored
disciplines such as labor economics, education, sociology, recommendations for candidates to choose the most suitable
and psychology. department. Compared to existing systems, the proposed
system provides more precise recommendations. To assess
In the contemporary era, selecting a profession isn’t merely candidate skill sets and determine the best department fit,
about earning a living; it’s about finding fulfillment and the system employs the K-Nearest Neighbor algorithm.
dedicating one’s time to something meaningful. Many Additionally, it utilizes the K- Means Clustering algorithm
individuals find themselves in jobs that fail to satisfy their to form clusters based on student scores in specific skill
skills and interests, highlighting the need to revamp sets, analyzing success rates across various departments
society’s conventional approach to career choices. To within each cluster.
facilitate better self-assessment and minimize dilemmas, an
online portal has been established for students. Rather In reference [1], a Career Guidance System specifically
than blindly selecting a career path without understanding designed for students studying Science in the 11th and 12th
one’s strengths, which often leads to frustration and grades is put forth. This system aims to assist students in
disillusionment, it’s prudent to embrace self-discovery. selecting suitable career paths post-12th, aligning with their
This initiative aims to empower students to make informed aptitude, emotional quotient, personality traits, and
decisions based on their unique abilities and aspirations, interests. The accuracy rates for the employed algorithms
free from peer influence or impulsive choices. The system are as follows: The K Nearest Neighbor algorithm yields a
caters specifically to students interested in pursuing careers classification accuracy of 65%, while Random Forest
in information technology and informatics, offering achieves 81%, and Naive Bayes demonstrates a
comprehensive assessments covering pertinent topics and classification accuracy of 57%.
subjects in these fields.
The suggested system forecasts suitable career paths for
students in the 10th and 12th grades by utilizing
pertinent input data. Additionally, it offers a quiz feature in
Volume 13 Issue 6, June 2024
Fully Refereed | Open Access | Double Blind Peer Reviewed Journal
www.ijsr.net
Paper ID: SR24612203820 DOI: https://dx.doi.org/10.21275/SR24612203820 1140
 International Journal of Science and Research (IJSR)
ISSN: 2319-7064
SJIF (2022): 7.942
the student’s field of interest, allowing them to assess their the validation is completed ina skill-wise way.
knowledge. Accessible as both an Android and web
application, students can conveniently utilize the system on 3. Proposed System
desktops and mobile devices. Prediction tasks are executed
using the Na¨ıve Bayes Algorithm, incorporating a dataset This article introduces a prediction model through a web-
encompassing various at- tributes such as personal, based portal, primarily focusing on the CS/IT field. The
curriculum, and background details, as well as interests. By model evaluates the most straightforward career trajectory
providing tailored guidance based on individual interests for an individual based on a survey filled out by users. The
and skills, this application aims to empower students in model’s primary predictive algorithm, The KNN algorithm
making informed career decisions. leverages advanced technologies like Piston, complemented
by machine learning methodologies including Decision
The Career Guidance and Recommendation System, as Trees, Random Forests, and Support Vector Machines. By
proposed in [2], seeks to anticipate fitting career options for leveraging LinkedIn profiles, job roles have been mapped
students studying in the 10th and 12th grades by according to the requisite skills for specific positions,
utilizing pertinent input data. Additionally, it offers a enhancing the accuracy of our forecasts to align more
quiz feature in the student’s field of interest, enabling closely with actual requirements. Upon entering the portal,
them to assess their knowledge. Accessible as both an students are greeted with a welcome screen and landing page
Android and web application, students can conveniently providing a concise overview of the site. The subsequent
utilize the system on desktops and mobile devices. sections detail the two main operational steps of the
. platform.
A dedicated online platform has been established to Step 1: The first stage involves collecting input from
aggregate data from applicants, conducting assessments on students, who answer questions related to their personal
their aptitude, interests, and personality traits. This skillsand academic grades.
dataset serves as the training set for the Machine Learning Step 2: Job Role Suggestion
model, which employs advanced algorithms to generate Upon submission of their responses, students receive career
optimal career path predictions based on the input recommendations, typically up to a maximum of three top
variables. job roles. They are also given the option to explore further
details regarding the provided predictions.
As Suggested in Paper [11], refinement of hyperparameters
and improvements made to the model, The Student Career
Guidance System, incorporating Machine Learning as
delineated in [4], deploys the Logistic Regression model
to achieve a 74% accuracy rate. While adopting
advanced machine learning classification techniques like
Decision Trees, Gradient Boosting, and Support Vector
Machines may yield more refined models, this endeavor
necessitates a substantial expertise level owing to the
comprehensive array of features integrated into the model.

1) QUIZ’S based on Skill

As Described in [9] The candidate completes an assess-
ment in this module that consists of both psychological and
fundamental skill-oriented questions. HTML 5, CSS 3, and
JavaScript are only a few of the web technologies that were Figure 1: System Block Diagram
used in the development of this module. HTML version 5,
often known as Hyper Text Markup Language 5, is a 4. Methodology Used
markup language used to create pages that shown within a
computer browser. It provides the document with a skeleton Data collection stands as one of the most crucial tasks in
structure; if we only use HTML, the document will be static. any machine learning research, as it serves as the input for
A style sheet language called Cascading Style Sheets 3 algorithms. Consequently, the accuracy and efficiency of
(CSS version 3) is specifically used to add styles and these algorithms hinge greatly on the quality and precision
designs to HTML documents in order to improve their of the collected data. In this study, two datasets, denoted as
display. The client-side scripting language is called [5]and [6], were merged to form a unified dataset
JavaScript. utilized in the

As Suggested in Paper [8] JavaScript is essential to back-

end programming, whereas HTML 5 and CSS 3 are
essential to front-end development. There will be multiple
choice answers for each question displayed separately.
JavaScript will be used to help with the validation process,
where each option in a question will have a different
weight based on which answer is the most appropriate. The
final result will be provided according to skill level after

Volume 13 Issue 6, June 2024

Fully Refereed | Open Access | Double Blind Peer Reviewed Journal
www.ijsr.net
Paper ID: SR24612203820 DOI: https://dx.doi.org/10.21275/SR24612203820 1141
 International Journal of Science and Research (IJSR)
ISSN: 2319-7064
SJIF (2022): 7.942
consistently sustains an accuracy exceeding 90%.

This was achieved using a dataset meticulously crafted

through a combination of diverse resources and
methodologies.

Figure 2: Algorithms Used

prediction model. One dataset encompasses student

interests, while the other contains student marks.

Dataset Pre-Processing: The two datasets initially held

disparate value types; one featured numeric values within
ranges, while the other comprised numeric values Figure 4: Random Forest
represented as percentages. To align them for compatibility,
the dataset was normalized, converting the values into As Described in [12] Random Forest classifier is composed
categorical data. of numerous decision trees created from various subsets of
a given dataset. Its predictive accuracy is improved through
Dataset Training and Testing: For our model, the training the aggregation of outcomes across these trees.
and testing data are partitioned in a 70:30 ratio to facilitate
effective training and evaluation. Diverging from dependency on an individual decision tree, a
random forest amalgamates forecasts generated by
Algorithms used for prediction: numerous trees, employing a majority voting mechanism to
As Described in [10] For prediction, the algorithms utilized ascertain the ultimate outcome. Augmenting the forest’s tree
include the straightforward technique called K-Nearest count not only enhances predictive accuracy but also serves
Neighbor (KNN). KNN is employed to determine the to alleviate the potential for oversampling.
unknown function with adequate accuracy, relying on local
minima of the objective function. It calculates the range A decision tree as suggested in [3] serves as a supervised
or distance of the unknown input using metrics such as learning technique applicable to both classification and
Euclidean distance, Minkowski distance, cosine similarity regression tasks, although it’s primarily recommended for
measure, chi-square,and correlation properties. classification purposes. Structured like a tree, it consists of
nodes that represent features of the dataset, branches
Of all the modules, the prediction module holds paramount indicating decision rules, and leaf nodes representing
importance and serves as the foundation. Various outcomes. The tree includes two key types of nodes:
technologies were leveraged in the development of this decision nodes, facilitating decisions with multiple
module, including machine learning techniques, APIs, and branches, and leaf nodes, These trees exclusively consist of
datasets for modeling and training. As Described in [13], terminal nodes devoid of further branches. Decisions or
Python serves as the primary programming language across evaluations within the tree are contingent upon the attributes
all applications. present in the dataset, and it furnishes a visual
representation delineating all conceivable solutions or
decisions contingent upon specific conditions.

Support Vector Machine (SVM) in [15], stands out as a

widely employed supervised learning algorithm applied to
address classification and regression tasks, though its
primary utility lies in machine learning classification
endeavors. SVM focuses on identifying extreme points or
Figure 3: KNN vectors to construct the hyperplane. These extreme
instances, termed support vectors, delineate the essence of
As Suggested in [7]K-Nearest Neighbor (KNN) excels in the algorithm, hence its designation as a support vector
handling classification problems, showcasing its machine. Consider the
proficiency. Despite the utilization of alternative
classification algorithms like support vector machine,
random forest, and decision tree, K-nearest neighbor
Volume 13 Issue 6, June 2024
Fully Refereed | Open Access | Double Blind Peer Reviewed Journal
www.ijsr.net
Paper ID: SR24612203820 DOI: https://dx.doi.org/10.21275/SR24612203820 1142
 International Journal of Science and Research (IJSR)
ISSN: 2319-7064
SJIF (2022): 7.942
Table I: Accuracy Table
Classification Techniques Accuracy
KNN 0.9410
Decision Tree 0.8906
SVM 0.8632
Random Forest 0.8105

graph below with two different classes classified using a

decision boundary or hyperplane. In the context of a binary
classification scenario, the goal of SVM is to find a
hyperplane that separates at most two classes. This
hyperplane acts as a decision boundary, and the support
vectors are the instances closest to this boundary.

We tested out a number before selecting the one that

produces the best result using the gridsearch technique. In
order to determine the combination of hyperparameter
values that yields the greatest performance for a particular
model, GridSearch entails creating a grid and doing a
methodical search over it. The most common method for
doing the search is cross-validation, in which the dataset is
partitioned into several subsets, or folds, and the model is
trained and assessed several times, using a distinct subset
for validation each time.

5. Result
Figure 6: Quiz Assessment
This online portal provided comprehensive
recommendations to students using the information The innovative quiz is meticulously crafted to gauge
gathered from their responses during the assessment aptitude and inclination across diverse disciplines such as
process. Predictive output is obtained by mapping these computer science, mechanical engineering, civil
responses to relevant job categoriesand providing users with engineering, and more. By leveraging the power of
personalized career recommendations. technology, this screening tool aims to provide a
comprehensive understanding of an individual’s preferences
Our primary mission is to help students identify and and strengths, thereby aiding in career guidance and
understand their unique gifts, personalities and talents. educational counseling. The screenshot showcased in the
With the innovative career guidance system built into our paper encapsulates the user interface of the quiz, illustrating
portal, people can make informed decisions about their its user-friendly design and the array of questions
career. The system becomes a guiding force that not only formulated to delve into the intricacies of one’s interest in
helps in choosing a career, but also empowers users to direct distinct fields. This research marks a significant stride in the
their choices according to their true interests. The focal realm of personalized career assessment, fostering a data-
point of our project driven and technology- driven approach to guide individuals
towards fulfilling and compatible professional paths.

Figure 5: Suggestion Page Figure 7: Prediction result

lies in providing career recommendations to students. We’ve been successful in developing an online portal that
Through our project, we have effectively delivered provides suitable recommendations to aid candidates in
personalized career guidance to students. The prediction is choosing the ideal position. We have applied the K-Nearest
generated based on the responses provided during the Neighbor algorithm for categorization. Assisting students in
assessment, where the answers are mapped to relevant job realizing their skills, personalities, and talents is the main
categories to derive the outcome. objective of the initiative. Students can use a career
advising system to help them choose a professional path.
It will also allow them to follow their passions. This

Volume 13 Issue 6, June 2024

Fully Refereed | Open Access | Double Blind Peer Reviewed Journal
www.ijsr.net
Paper ID: SR24612203820 DOI: https://dx.doi.org/10.21275/SR24612203820 1143
 International Journal of Science and Research (IJSR)
ISSN: 2319-7064
SJIF (2022): 7.942
approach helps graduates in the CS/IT stream select a Douglas C. Schmidt, editors, On The Move to
career. Meaningful Internet Systems 2003: CoopIS, DOA, and
ODBASE, pages 986–996, Berlin, Heidelberg, 2003.
6. Conclusion Springer Berlin Heidelberg.
[8] Simon Holm Jensen, Anders Møller, and Peter
Incorporating predictive algorithms into job opportunities Thiemann. Type analysis for javascript. In Jens
adds a forward-looking dimension, enabling users to Palsberg and Zhendong Su, editors, Static Analysis,
proactively develop their skills and aspirations. The pages 238–255, Berlin, Heidelberg, 2009. Springer
seamless integration of applying directly through our portal Berlin Heidelberg.
not only simplifies the job search process, but also makes [9] Atefeh Khosravirad, Ahmad Akbari, Mohammad
the transition from skill acquisition to practical application Karimi, and Moslem Cherabin. Designing a skill-
easier. We’ve developed a user-centric platform that not based education model in schools based on the real
only addresses im- mediate career needs but also cultivates needs of society using grounded theory. Journal of
a supportive community and encourages ongoing growth Adolescent and Youth Psychological Studies (JAYPS),
and progress. Alongside providing a chat room for direct 5(2):1–13, Feb. 2024.
communication and offering skill-specific courses tailored [10] Vivekkumar Mourya, Shubham Phatale, Siddhesh
to various job requirements, we aim to expand our Thakur, and Panjab Mane. Career guide application
integration to encompass all relevant departments. Our using ml. International Research Journal of
objective is to incorporate platforms like LinkedIn, Udemy, Engineering and Technology (IRJET), 7(09), 2020.
and others into our services, facilitating connections [11] Heungseok Park, Yoonsoo Nam, Ji-Hoon Kim, and
between users and experts in their respective fields. This Jaegul Choo. Hyper- tendril: Visual analytics for user-
integration offers valuable networking opportunities and driven hyperparameter optimization of deep neural
advice to support users’ development and enhance their networks. IEEE Transactions on Visualization and
chances of success in their chosen careers. Our role is to Computer Graphics, 27(2):1407–1416, 2021.
showcase potential positions predicted by our algorithm, [12] Aakash Parmar, Rakesh Katariya, and Vatsal Patel. A
enabling candidates to promptly submit applications review on random forest: An ensemble classifier. In
through our portal. Jude Hemanth, Xavier Fernando, Pavel Lafata, and
Zubair Baig, editors, International Conference on
References Intelligent Data Communication Technologies and
Internet of Things (ICICI) 2018, pages 758–763,
Cham, 2019. Springer International Publishing.
[1] Dahanke Ajay, Shinde Nilesh, Dhagate Anirudh,
[13] KR Srinath. Python–the fastest growing programming
Shaikh Huzaif, and PV Nagare. An intelligent career
language. Interna- tional Research Journal of
guidance system using machine learning. 2022.
Engineering and Technology, 4(12):354–357, 2017.
[2] Hmood Al-Dossari, Fawaz Abu Nughaymish, Ziyad
[14] S Vignesh, C Shivani Priyanka, H Shree Manju, and
Al-Qahtani, Mo- hammed Alkahlifah, and Asma
K Mythili. An intelligent career guidance system
Alqahtani. A machine learning approach to career path
using machine learning. In 2021 7th International
choice for information technology graduates.
Conference on Advanced Computing and
Engineering, technology & applied science research,
Communication Systems (ICACCS), volume 1, pages
10(6):6589–6596, 2020.
987–990, 2021.
[3] Bahzad Charbuty and Adnan Abdulazeez.
[15] S.V.M. Vishwanathan and M. Narasimha Murty.
Classification based on decision tree algorithm for
Ssvm: a simple svm algorithm. In Proceedings of the
machine learning. Journal of Applied Science and
2002 International Joint Conference on Neural
Technology Trends, 2(01):20 – 28, Mar. 2021.
Networks. IJCNN’02 (Cat. No.02CH37290), volume
[4] Ankush Daharwal, Sandeep Gore, Aishwarya
3, pages 2393–2398 vol.3, 2002.
Bhagwat, Shraddha Dethe, and Sunny Chavan. Career
guidance system using machine learning for
engineering students (cs/it). 2020.
[5] Samruddhi Deshpande, Pooja Gupta, Nikita Singh,
and Deepali Kadam. Prediction of suitable career for
students using machine learning. Inter- national
Research Journal of Engineering and Technology,
8(02), 2021.
[6] Serhii Dolhopolov, Tetyana Honcharenko, Svitlana
Anastasiia Dol- hopolova, Olena Riabchun, Maksym
Delembovskyi, and Oksana Omelianenko. Use of
artificial intelligence systems for determining the
career guidance of future university student. In 2022
International Conference on Smart Information
Systems and Technologies (SIST), pages 1–6. IEEE,
2022.
[7] Gongde Guo, Hui Wang, David Bell, Yaxin Bi, and
Kieran Greer. Knn model-based approach in
classification. In Robert Meersman, Zahir Tari, and
Volume 13 Issue 6, June 2024
Fully Refereed | Open Access | Double Blind Peer Reviewed Journal
www.ijsr.net
Paper ID: SR24612203820 DOI: https://dx.doi.org/10.21275/SR24612203820 1144