International Journal of Scientific Research in Engineering and Management (IJSREM)

Volume: 07 Issue: 06 | June - 2023 SJIF Rating: 8.176 ISSN: 2582-3930

Screening and Ranking Resume Using Stacked Model

Guide: - Prof. Masira Kulkarni
GHRIET, PUNE

Rushikesh Patil Ashish Raut

Department: Computer Engineering Department: Computer Engineering
GHRIET PUNE GHRIET PUNE
[email protected]
[email protected]

Pawan Pawar Prasad Pente

Department: Computer Engineering Department: Computer Engineering
GHRIET PUNE GHRIET PUNE
[email protected] [email protected]

ABSTRACT posts a job offer. Every recruiter's first step in

The goal of the resume screening process is to find the hiring process is to screen resumes. A
the best candidates for a vacancy. The program company's HR department will likely get tens
must employ both natural language processing and of thousands of resumes and cover letters per
machine learning techniques to rate applicants in day if there is a position open.
real time. Natural language processing (NLP) and
Since it requires human engagement, this
machine learning algorithms are used as methods
method currently takes the longest and is the
to rate resumes. The result would be a top
most error prone. Humans are limited in ways
candidate's resume, with resumes and job
that make it impossible for them to work
descriptions acting as the input. The output's continuously. Hence, decreasing efficiency is
findings are immediately available. Mong will be also a problem. Therefore, to circumvent the
used for TF-IDF, Cosine Similarity, and string company's restriction, we have proposed a
matching. Although current methods are simple system that could swiftly determine the
and effective, they fall short in terms of precision, required skill set by examining a document or
efficiency, and computing power. even a résumé and sorting the outcomes
according to skill sets. One tactic we want to
Keywords use is machine learning. Recruitment
Deep Learning, Natural Language technology is a developing area of HR
Processing, Machine Learning, and TD-IDF technology that intends to automate or
1 INTRODUCTION considerably streamline formerly manual
The worldwide recruiting market is estimated procedures like resume screening.
to be worth $200 billion. To locate the New research is conducted daily in a huge
candidates who are most suited for a certain number of universities across many different
job description, a large pool of applications fields. Everyday living benefits from a vast
must be combed through. A company will get array of research and discoveries, from the
dozens of applications for the position when it

© 2023, IJSREM | www.ijsrem.com | Page 1

 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 07 Issue: 06 | June - 2023 SJIF Rating: 8.176 ISSN: 2582-3930

field of information technology to that of The best similar resumes are then provided to
medicine. Due to these changes, many the recruiter.
industries today have more opportunities
accessible than ever before. It is reasonable to 2 Literature Review
suppose that similar trends exist in other The best applicants are assessed in line with
countries' hiring procedures. Although the the job description in (1) using K-NN,
resume serves as the foundation of the Content-based Recommendation, and cosine
applicant's application, it is the candidate who similarity to identify the CVs that match the
must build it in the beginning. job description the most closely.

Most of the contemporary employers’ demand In (2), the proposed system, JARO, is a virtual
that candidates submit their resumes online. assistant that conducts interviews by
When they have gotten the email, the next step reviewing the résumés of candidates (CVs). A
is to group the replies according to the list of inquiries to ask the requestor will then
situation. Usually, they sort this by hand, but it result from this. To ensure an objective
takes a lot of time. On the other hand, it's judgement, this expedites the interview
possible that many errors are brought on by it. process.
This type of sorting is rather ineffective. This
To determine the applicability of each resume,
might result in employing someone whose
(3) uses a neural-network model and a few
credentials are insufficient for the post or it
text-processing techniques. We provide a
could result in the manual sorting procedure
variety of model solutions that either make use
missing a great candidate.
of the pair/triplet-based loss or a semi-
The most time-consuming aspect of a supervised learning method to handle the label
recruiter's job is selecting the candidates from deficiency problem in the dataset.
a large pool who are best qualified for the
Inside (4), multiple regression was employed
vacant position. For human recruiters to go
to evaluate the idea. The findings demonstrate
through hundreds of resumes to locate the top
how greater usage of AI in working contexts
candidates might be daunting when a company
increases HR functional performance by
gets them. There might be hundreds of
demonstrating a substantial positive
candidates for each given position at the
association between two criteria.
company. 75 percent of advertisements don't
show the necessary competence for the job The research of a hybrid technique for
being advertised on average. curriculum vitae screening conducted by the
authors was detailed in (5). An integrated
Because of LinkedIn, Facebook, and online
knowledge base is used to carry out the
talent databases, thousands of candidates may
classification job in this case and to
apply for a single post. Automated systems are
empirically demonstrate that they get
starting to prevail in modern HR departments.
promising accuracy results when compared to
The field is streamlined and improved with
conventional machine learning-based resume
resume screening. Hence, human resources
classification algorithms utilizing a real-world
professionals have more time to engage in
recruiting dataset.
fruitful conversations with both potential
workers and candidates. Our suggested design KNN, Linear SVC, and XGBoost are only a
is a digital tool for HR departments that few of the machine learning techniques
effectively screens resumes of applicants using employed in system (6). To accurately forecast
NLP technology. The two best methods for specific job profiles from a text description, all
comparing the information on resumes in the of these algorithms are merged into a two-
field to that on job descriptions are token- level stacked model. The usage of this
based similarity and latent Dirichlet allocation. approach may help both employers, waitlist

© 2023, IJSREM | www.ijsrem.com | Page 2

 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 07 Issue: 06 | June - 2023 SJIF Rating: 8.176 ISSN: 2582-3930

competitors and job seekers determine The system flow diagram shows how the
whether their resumes are sufficiently suggested model works. First, the input is
organized for the system to identify the right provided as research in PDF format, and then
job profiles. lexical and semantic analysis is conducted as
part of the preprocessing. Genism, which
Job searchers can submit their resumes and examines similar words and calculates a score
apply for any positions via the interactive web using LDA, produces a resume and a job
application in (7). NLP and machine learning description. This entire procedure contributes
are used by the corporate recruiter to compare to the creation of the web interface design.
resumes to job requirements.
In (8), it's online tool helps us sort and rank
resumes. The uploaded resumes are compared 3 Research Methodology
to the job description using a tool called The objective of the research is to identify
Natural Language Processing to determine available information communication technology
which ones are best suited for the position infrastructure in the schools for implementing the
(NLP). Then, the resumes are graded and rated information management system. The description
in order of best match to worst match. The study was conducted for implementation of
only people with access to this ranking are management information system.at schools. The
area of study was Hyderabad division primary and
corporate recruiters who are trying to select
secondary schools. Public and Private schools
the best candidates from a large pool of were respondents of the research. Questionnaires
applicants. were used as an instrument for data collection. The
first part of the questionnaire is regarding
In (9), The authors hope to automate this demographic information of respondents. 2nd part
screening process by analyzing NLP in depth of questionnaire facilities available for the
to change the entire recruitment process by implementation of information system and options
making employer responsibilities easier than of managers/principals regarding the technological
ever before, and we are capable of scoring facilities. Questionnaires have most questions are
individual CVs by comparing total fields in liker scale types from strongly agree, agree,
CVs with needed fields. Furthermore, their neutral, disagree and strongly disagree. The
classification model has an accuracy of 74% in population of study was mangers/ principals of the
terms of precision. schools who are working as an administrator. This
system can integrate all the media materials into
The writers compile resume-related data and the systematic resources, such as an entire school-
do the necessary analyses on it to provide based web-course or theme website. Before this
content that recruiters may utilize (10). The work, rural teachers should make good
Resume Parser would help recruiters find the instructional design on the web course, and upload
the media materials into the resource library, and
most qualified candidates rapidly, saving them then they can conduct the web course
time and effort. development. The whole process of developing a
school-based web-course mainly includes four
System Flow Diagram
steps: creating new courses, adding and managing
course catalog, adding and managing course
content, approving and publishing course content.

© 2023, IJSREM | www.ijsrem.com | Page 3

 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 07 Issue: 06 | June - 2023 SJIF Rating: 8.176 ISSN: 2582-3930

3. EXISTING SYSTEM 5. ALGORITHM COMPARISON

Nowadays, resume screening is done manually
through the evaluation of applications. It is
physically challenging to go through the
resumes of the millions of job seekers in
India's enormous employment market to find
someone who is a suitable match. This results
in a labor hiring procedure that is slow and
ineffective, which costs businesses resources.
HR must personally sift through each résumé
to find the applicant whose qualities most
closely match those listed in the job
description. There is a chance that the best
applicant for the job will be missed during the
We evaluate the performance metrics of the
selection process because this uses a lot of
two models and find the accuracy rates. It is
resources and is prone to mistakes.
found that NLP outperforms Navies Bayes.
4. PROPOSED SYSTEM
6. REFERENCES
Our approach uses machine learning and
natural language processing (NLP) to rank This forecast enables the resume to be selected
resumes. With the help of artificial without a manual review. It helps to improve
intelligence, this tool for contextually accuracy, and in the long run, this model itself
assessing resumes goes beyond keywords. educates the system to evaluate and choose the
After reviewing resumes, the software necessary abilities from the resume that is
evaluates prospects in real-time in accordance provided as input. As a result, prediction
with the hiring needs of the recruiter. The improvement is strong and efficient.
online application seeks to organize the
resumes by comparing the resumes that fit the 1. Pradeep Kumar Roy; Sarabjeet Singh
specified Job Descriptions and intelligently Chowdhary; Rocky Bhatia, "A Machine
reading job descriptions as input. After Learning technique for automation of Resume
filtering, it offers a ranking and suggests the Recommendation system," Procedia Computer
best resume for a specified text job Science Volume 167, Pages 2318-2327, 2020
description. The software uses natural
language processing to instantly match and 2. Jitendra Purohit; Aditya Bagwe; Rishabh
rate applicants. When it comes to string Mehta; Ojaswini Mangaonkar; Elizabeth
matching, cosine similarity, overlap George, "Natural Language Processing based
coefficient, and natural language processing, Jaro-The Intervieweing Chatbot," 2019 3rd
this software uses Mong rather than general International Conference on Computing
methods. Methodologies and Communication (ICCMC),
We employ a different methodology for our August 2019
work because it largely concentrates on the
content of the resumes, from which we extract 3. Yong Luo, Huaizheng Zhang, Yongjie
skills and other pertinent information to match Wang, Yonggang Wen, and Xinwen Zhang,
people with job descriptions. "ResumeNet: A Learning-Based Framework

© 2023, IJSREM | www.ijsrem.com | Page 4

 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 07 Issue: 06 | June - 2023 SJIF Rating: 8.176 ISSN: 2582-3930

for Automatic Resume Quality Assessment,"

2018 IEEE International Conference on Data
Mining (ICDM), December 2018
[ResumeNet: A Learning-Based Framework
for Automatic Resume Quality Assessment]

4. 2020 International Conference on

Computation, Automation, and Knowledge
Management (ICCAKM): Garima Bhardwaj,
S. Vikram Singh, and Vinay Kumar, "An
Empirical Study of Artificial Intelligence and
Its Impact on Human Resource Functions."

5. Mohamed Maree, Muath Sabha, and Abeer

Zaroor's book chapter from 2017's Smart
Innovation, Systems and Technologies volume
72

6. Ranikka Ransing Nikita

Bhrugumaharshi Emberi, Kailas
Mahavarkar, and Akshaya Mohan,
"Screening and Raking resumes based on
Stacked Model, 2021, IEEE

7. Amin Sujit Web Application For

Screening Resumes, Nikita Jayakar, Sonia
Sunny, Pheba Babu, M. Kiruthika, and
Ambarish Gurjar, 2019 IEEE.

8. M. Alamelu D.Sathish Kumar; R.

Sanjana; J.Subha Sree; A.Sangeerani
Devi; D. Kavitha “Resume Validation and
Filtration using NLP, 2021, IEEE
IEMECON

9. M.F. Mridha; Rabeya

Basri; Muhammad Mostafa Monowar; Md.
Abdul Hamid “A Machine Learning
Approach for Screening Individual’s Job
Profile Using Convolutional Neural
Network”, 2021 IEEE

10. Anushka Sharma, Smiti Singhal, and

Dhara Ajudia, 2021 AIMV, Artificially
Intelligent and Mechanics Voices

© 2023, IJSREM | www.ijsrem.com | Page 5