IAES International Journal of Artificial Intelligence (IJ-AI)

Vol. 13, No. 4, December 2024, pp. 4334~4343

ISSN: 2252-8938, DOI: 10.11591/ijai.v13.i4.pp4334-4343  4334

Eligibility rate of applicant’s LinkedIn account: a naïve bayes

classification and visualization

Khyrina Airin Fariza Abu Samah1, Nurul Athirah Ahmad1, Anis Amilah Shari1,
Hana Fakhira Almarzuki2, Zuhri Arafah1, Lala Septem Riza3, Amir Haikal Abdul Halim1
1
College of Computing, Informatics and Mathematics, Universiti Teknologi MARA Melaka Branch, Melaka, Malaysia
2
College of Computing, Informatics and Mathematics, Universiti Teknologi MARA Shah Alam, Selangor, Malaysia
3
Department of Computer Science Education, Universitas Pendidikan Indonesia, Bandung, Indonesia

Article Info ABSTRACT

Article history: In the digital era, social media platforms like LinkedIn have become famous
for recruitment, and recruiters widely use them to find potential employees.
Received Dec 22, 2023 The recruitment process is crucial in organizations, as it involves selecting
Revised Mar 13, 2024 qualified applicants from a diverse pool. However, the screening process and
Accepted Mar 21, 2024 manual recruitment process entail significant time, high costs, and potential
bias. Consequently, it may cause recruiting unqualified applicants and may
affect the organizations. Thus, this study aims to classify and generate a list
Keywords: of potential job applicants by analyzing seven attributes of their LinkedIn
accounts: title, location, skills, education, language, certification, and years of
Classification experience. Data are collected from LinkedIn profiles and then undergo data
LinkedIn pre-processing. The naive Bayes (NB) algorithm is implemented as the
Naïve Bayes classification algorithm and sets the classification as “eligible” or “ineligible”.
Recruitment The NB model achieved an accuracy testing of 89.8%, indicating good
Visualization performance in classifying potential job applicants. At the same time, we
measure the similarity cosine score to set the mean of the eligibility. The
classification results are visualized for the suitable applicants in descending
rank, allowing users to choose the applicants’ classification status efficiently.
For the system usability, we managed to get 90% from the recruitment expert.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Khyrina Airin Fariza Abu Samah
College of Computing, Informatics and Mathematics
Universiti Teknologi MARA Melaka Branch, Jasin Campus, Merlimau, 77300, Melaka, Malaysia
Email: [email protected]

1. INTRODUCTION
Recruitment is a crucial aspect of the human resources (HR) Department as it involves selecting
eligible candidates from a vast applicant pool [1]. Employing a variety of strategies is necessary to locate,
interview, and recruit individuals for the position. The HR Department’s initial task is recruitment to ensure
each employee is competitive and contributes to society [2]. Interviewers aim to identify the most suitable
candidate who fulfils the employment requirements during this process. Abbas et al. [3] emphasized the
significance of selecting the appropriate candidate in company operations, stating that competent employees
can significantly impact the success or failure of an organization. Lawong et al. [4] stated that both the
organization and the agents share responsibility for the effectiveness of the hiring process. Recruiters play a
crucial role in an organization’s performance by implementing effective hiring and recruiting tactics to attract
qualified and competent individuals. They accomplish this by conducting research, designing, and
implementing these tactics.

Journal homepage: http://ijai.iaescore.com

Int J Artif Intell ISSN: 2252-8938  4335

In modern times, the world has undergone a digital transformation [5], including social media
platforms such as Facebook, LinkedIn, and Twitter, which are more popular for recruitment processes [6].
LinkedIn is the primary site for companies to recruit applicants and emphasizes cultivating professional
connections [7], [8] with a usage rate of 77% compared to other platforms [9]. It is the biggest online
networking site for professionals, linking more than 900 million individuals in over 200 countries. It enables
users to discover job openings, broaden their professional connections, and acquire new skills to achieve
success in their careers. According to Wei [10], Malaysia has around 5.79 million LinkedIn users, and
numerous recruiters prefer it for recruitment purposes [11].
One step in the recruitment process is the screening of applications [12]. It involves evaluating job
candidates to assess their suitability for a position. However, Sivanandam and Mudaliar [13] highlighted that
recruiters find it challenging to go through numerous resumes. Examining the applicant’s resume is a time-
consuming process, leading to delays and ineffective time management [14]. Recruiters must verify and
evaluate the minimal credentials for the job to guarantee a successful recruitment process and make an informed
selection. Abbas et al. [3] stated that a screening process that is not effective could result in generating a roster
of inadequately qualified candidates.
Moreover, a manual recruitment method necessitates substantial expenses [15], [16]. Costs and
expenses related to recruitment must be considered, such as time to hire, resume screening, and recruiter fees.
Furthermore, the expenses would be at their highest if unqualified candidates were selected for the position.
Utilizing inexpensive recruitment processes may result in erroneous shortlisted candidates as they do not
always ensure the most eligible applications [17]. A manual recruitment procedure might lead to biased
outcomes influenced by gender or human perception, affecting the decision-making in the recruitment process
[18], [19]. Recruitment bias occurs when the recruiter assesses the applicant only based on their initial
impression. It is influenced by human perception, making individuals more inclined to favour a resume with
an appealing profile image.
Eight high-demand vocations in Malaysia for 2022 have been determined based on current trends and
industry estimates [20]. The list comprises information technology (IT), software development, digital
marketing, finance positions, project management roles, business development and sales executives, medical
professionals, educators, and customer service executives. IT and software development, digital marketing,
and finance jobs are the top three jobs identified by JobStreet and chosen for this research. Job seekers should
prioritize reviewing the job requirements information that can provide insight into the specific job requirements
for recruiters [21], [22]. Attributes listed in job listings are crucial for recruiting appropriate candidates.
Researchers identified 10 key factors for recruiters to consider when evaluating a job opportunity: position,
skills, education level and history, languages, years of experience, certification, salary, benefits, location, and
working hours [23], [24]. However, this research focuses on seven attributes for classification: title, location,
education, years of professional experience, skills, languages, and certificates [25], [26].
A web-based dashboard was created utilizing data extracted from LinkedIn profiles that were scraped
from the platform in response to the identified issues. The naïve Bayes (NB) algorithm was utilized to classify
and visualize the LinkedIn accounts of applicants who meet the company’s job requirements. The system
utilizes bar charts and pie charts for visualization. It allows users to see which applications from their LinkedIn
accounts meet the requirements for the job. It assists recruiters in identifying the most suitable candidate who
meets the job’s criteria. The paper is structured as follows: section 1 commences with a concise introduction.
Section 2 details the approach, while section 3 presents the results and discussion. Section 4 ends the study and
offers a brief review of potential future improvements.

2. RESEARCH METHOD
2.1. Design of the system
System design in research involves developing a framework or structure to investigate and solve a
specific research issue. We deliberated on the comprehensive system architecture, system flow, interfaces, and
the data pertaining to the system needs. During this stage, we utilize a use case diagram and a flowchart diagram
to illustrate the workflow. User interface (UI) is the ultimate stage in the design process. The term pertains to
the visual arrangement of the system components that a user can engage with on a website. UI design should
be effective and user-centric to guarantee user-friendliness and appeal to potential users. It strives to streamline
the user’s interaction to efficiently achieve their goals within the system.

2.2. Back-end development

Back-end development includes gathering, preparing, and applying the NB model, as seen in
Figure 1. The diagram illustrates the system’s operation from data collecting initiation to visualization
completion. Data preparation includes collecting and pre-processing data to guarantee its reliability, and then
using an NB classifier for categorization. The output consists of both visualized and reported data.
Eligibility rate of applicant’s LinkedIn account: a naïve bayes… (Khyrina Airin Fariza Abu Samah)
4336  ISSN: 2252-8938

Figure 1. Research design flow diagram for LinkedIn eligibility

2.2.1. Data collection and pre-processing

Data was collected by web scraping from September 1, 2022, to March 31, 2023, utilizing the
selenium and beautiful soup packages in Python. Selenium is a potent instrument for managing web browsers
and executing browser automation [27]. This study involves using selenium to automate the LinkedIn page to
extract applicant data. Subsequently, beautiful soup was utilized to retrieve the LinkedIn profile information.
The scraper navigates through all profile pages of the corresponding URLs on the LinkedIn website. The tool
retrieves LinkedIn profile information using specific parameters, including title, location, education,
experiences, certifications, skills, and languages. The extracted data was saved in a CSV format, and 2006
profiles were gathered for this investigation using scraping.
The data that was gathered was then cleaned up to make high-quality data for testing and training. It
was processed to remove symbols, fill in missing numbers, and lowercase all the letters in the dataset. It was
changed to the word “none” in the information where the empty list with the symbol “[]” used to be. After that,
a similarity score method was used to figure out how similar the applicant’s personal data was to the job
description data. The cosine similarity function was used to figure out the closeness score. This study looks at
titles, skills, and languages and gets the similarity score between them. The equation for cosine similarity score
is represented as in (1) [28].

(𝐴⋅𝐵)
𝐶𝑜𝑠𝑖𝑛𝑒 𝑆𝑖𝑚𝑖𝑙𝑎𝑟𝑖𝑡𝑦 = (||𝐴||⋅||𝐵||) (1)

Next, for education attributes, the education level was extracted using regex to match the education
level in that job’s requirements. If a match was found, it was assigned a value of “1” to indicate a successful
match in terms of education level. Afterwards, the duration of the applicant’s LinkedIn profile experiences was
computed in months to facilitate better comparisons with the minimum duration of experiences required by
each job’s specifications.
After all the attributes were processed, a few conditions were applied to the data labelled. A new
column called “eligibility status” has been added to the dataset. These eligibility status attributes have “eligible”
and “ineligible” statuses. The LinkedIn profile was classified as having “Eligibility” status if it fulfilled these
three conditions. First, the similarity score between title, skills and languages is higher than the mean of the
similarity score between title, skills, and languages. Function mean() calculated the average of the total
similarity score for the applicants and was commonly calculated as in (2). The second condition is if the
education level that was extracted earlier matches the education level required by the job’s description. The
last condition is that the duration of the LinkedIn profile experiences in a month should be higher than the
duration of the job experiences in a month. Otherwise, if the data does not meet those conditions, it will fall
under the “ineligible” category.

(𝐶𝑜𝑠𝑖𝑛𝑒_𝑆𝑖𝑚𝑖𝑙𝑎𝑟𝑖𝑡𝑦(𝐴1,𝐵1) +𝐶𝑜𝑠𝑖𝑛𝑒_𝑆𝑖𝑚𝑖𝑙𝑎𝑟𝑖𝑡𝑦(𝐴2,𝐵2) + …+
𝐶𝑜𝑠𝑖𝑛𝑒_𝑆𝑖𝑚𝑖𝑙𝑎𝑟𝑖𝑡𝑦(𝐴𝑁,𝐵𝑁))
𝑀𝑒𝑎𝑛 = (2)
𝑁

The final dataset has a total of 14 columns for LinkedIn profile: name, title, location, experiences,
education, certifications, skills, languages, similarityscores_title_skills_lang, education level, education match,
total duration in months, has certification, and eligibility. Finally, the dataset is saved in a CSV file for the next

Int J Artif Intell, Vol. 13, No. 4, December 2024: 4334-4343

Int J Artif Intell ISSN: 2252-8938  4337

training and testing to develop the model and used for data visualization. Once the pre-processing phase is
finished, the final dataset is prepared for the classification method using machine learning.

2.2.2. Naïve Bayes classification model

The NB classifier is a widely used algorithm for classification and prediction analysis [29]–[31]. NB
classifier is performed based on the Bayes theorem. It assumes that the features are conditionally independent
given the class label. The NB classifier calculates the probability of each class label given the input features.
It assumes that the features within the input features are conditionally independent given the class. By
leveraging this assumption, the NB algorithm can make predictions accurately and quickly. Bayes’ theorem
calculates the probability of an event happening based on the probability of another event that has already
occurred. According to Brownlee [32], the Bayes theorem was depicted as in (3).

𝑃(𝐵|𝐴)𝑃(𝐴)
𝑃(𝐴|𝐵) = (3)
𝑃(𝐵)

During training data, the NB classifier calculates the prior probability, which is P(A) from (3) of each
class, by counting the number of instances. It also estimates the conditional probabilities by calculating the
likelihood, P(B│A), of observing each feature value given the class label. Thus, during testing data, the Bayes
theorem was applied to calculate the posterior probabilities of each class given the input features, which is
P(A│B) from the equation. The class with the greatest posterior probability is subsequently chosen as the
predicted class. The dataset is labelled based on its conditions, where we use a ratio of 80:20 to split the data
into 2 parts: training and testing. Four feature columns were selected for training sets, such as similarity scores
title skills lang, education match, total duration in months, and has certification. After developing the NB
model, the eligibility rate was produced using Bayes’s theorem probability.

2.3. Front-end development

HTML, CSS, and JavaScript are three core technologies for developing and designing websites and
web applications. A website has been created for users to interact with, and the system can be accessed easily.
The web application connects the interface and the Python code using the Flask application. It displays the
result of the prediction and offers tools for data visualization, allowing the creation of charts and graphs to
represent data visually.

2.4. Testing development

Upon completion of all phases, the final phase is testing, which involves executing test cases on the
system. The verification and validation process ensures that the system fulfils the initial requirements and
satisfies its aims. Verification evaluates the system’s development product, whereas validation assesses the
final product. The testing phase is intended to ensure the system’s functionality and address any problems or
concerns. In this study, the objectives focus on dashboard functionality based on the defined use case and
system usability scale (SUS) by HR recruitment professionals.

3. RESULTS AND DISCUSSION

3.1. Model accuracy
Evaluating the NB classifier model involves comparing the predicted values with the actual values in
a test dataset to determine accuracy. Figure 2 displays the accuracy testing results of the NB model in this
study. The accuracy result is 89.8%. An accuracy score above 90% is considered “excellent” in [33], whereas
a score between 70% and 90% is classified as “good.” An accuracy score ranging from 60% to 70% is
considered “okay,” while a score below 60% is labelled as “poor.” The system’s accuracy ranges from 70% to
90%, which is considered “good.” The model achieved an accuracy rate of about 89.8% in classifying events
within the test dataset. The confusion matrix provides a detailed breakdown of the true values that align with
the model’s predictions. The confusion matrix in this study uses the classes “ineligible” and “eligible”, which
are denoted by “0” and “1” correspondingly.

3.2. Cosine similarity scores result

We use an example using the job’s field, “finance,” and the job’s title, “accountant,” to illustrate the
result. We obscured the original names with black markings. Then we renamed them using a comparable
alphabet to prevent any ethical concerns. We are deviating from the conventional alphabetical sequence
because the visualization is arranged according to the descending rate outcome. Cosine similarity scores are
computed using the titles, abilities, and languages of seven applicants. Applicants who exceed the average
similarity score are deemed to have met one of the eligibility criteria when the rules are applied. Thus,
Eligibility rate of applicant’s LinkedIn account: a naïve bayes… (Khyrina Airin Fariza Abu Samah)
4338  ISSN: 2252-8938

Figure 3 shows the bar chart of cosine similarity score comparisons where each bar represents an applicant’s
similarity score. A line with a dotted point represents the mean of the cosine similarity score of 0.26, which
indicates the mean of the similarity score across all applicants. In conclusion, only four candidates are eligible
since they surpass the mean based on the overall applicant’s score.

Figure 2. Naïve Bayes model Figure 3. Bar graph of similarity cosine score of applicants with
accuracy results mean score

3.3. Functionality testing of dashboard visualization

The dashboard visualization is related to functionality testing. It is the process where the UI is tested
by providing appropriate test input. The output is compared with the expected output. The system’s output will
be evaluated whether it is successful or not at the end of the test. This testing is mainly concerned with the
processing results. The goals are to test and run through all the system functions from the first to the last page
of the application. It ensures the application runs, follows the requirements, and displays the desired outcome.
It performs a series of test cases covering the application’s various functionalities. Thus, functionality testing
helps identify bugs, errors, or deviations from the intended behaviour.
Figure 4 shows the “upload file” page where users can select the job’s field and title and upload a
CSV file containing a list of URLs of the LinkedIn applicant’s account that the user wants to perform the
classification. Using the same example for the similarity coefficient score, we chose the job’s field as “finance”
and the job’s title as “accountant”. Once the button “upload” is clicked, the scraping process will start, and the
job requirements selected and the applicant’s information will be displayed on the next page, as illustrated in
Figure 5. On top of the dashboard, the job’s requirement is listed according to the seven attributes chosen for
this study. There is also information on the total of applicants that match the requirements.

Figure 4. “Upload file” page Figure 5. View the “display information” page

The system allows the user to click the “classification” button to show the applicant’s information
with its eligibility status, either “eligible” or “ineligible” on the right row, as depicted in Figure 6. The user
may choose to click “download all applicants” to see the details. The system exhibits the eligible applicants

Int J Artif Intell, Vol. 13, No. 4, December 2024: 4334-4343

Int J Artif Intell ISSN: 2252-8938  4339

along with their corresponding eligibility rates once the “view eligible applicants” button is selected as shown
in Figure 7. At the top of the page, the dashboard visualizes the total number of eligible applicants, which is 4.

Figure 6. View the “applicant’s eligibility status” page

Figure 7. View the “eligible applicants” page

The user has the choice to click the “download list” button to refer to the detailed information. The
analysis page was displayed after the user clicked the “overview of applicants analysis” button. Figure 8 shows
a pie chart of the total eligible and ineligible applicants. There are four eligible applicants at 57.1% and three
ineligible applicants at a percentage of 42.9%.
Following the pie chart, Figure 9 shows the bar chart of the eligibility rate of the applicants plotted in
decreasing order to ease the analysis. The y-axis denotes the eligibility rate, whereas the x-axis signifies the
applicant’s name. The chart is composed of bars, with each bar representing an applicant and its height
corresponding to that applicant’s eligibility rate. With that, the user can quickly identify and analyze the top
applicants with the highest eligibility rate.

Eligibility Rate of the Applicant(s)

94.58 92.98
100 87.30 85.23
80
Rate (%)

60
40
20
0.50 0 0
0

Name

Figure 8. Pie chart of eligible applicant(s) vs. Figure 9. Bar chart of the eligibility rate of applicant(s)
ineligible applicant(s)
Eligibility rate of applicant’s LinkedIn account: a naïve bayes… (Khyrina Airin Fariza Abu Samah)
4340  ISSN: 2252-8938

The next page is the “attribute analysis” page. This page visualizes three proportions of total
applicants: eligible applicants and ineligible applicants. For each proportion, it analyses seven attributes used
for the classification in this research. There will be two visualizations in total to differentiate the classification.
The first analysis was visualized using a stacked bar chart, where each applicant’s name and
satisfaction status for each attribute were displayed. Figure 10 shows that out of four eligible candidates, Winda
fulfils the attributes of certification, experience, and education. Catriona and Kiroshini fulfil attributes for
certification and education, while Harshimah only fulfils education attributes.
The second analysis was visualized based on states in Malaysia, as in Figure 11. Figure 11(a) using
the pie chart of the total applicants based on the geographical distribution of applicants across different states
in Malaysia: WP Kuala Lumpur, WP Putrajaya, Selangor, and Negeri Sembilan. It gives an overview of which
states have a higher number of applicants. Then, we split the pie chart in Figure 11(b) to determine which state
the four eligible candidates are. As can be seen, there are two eligible candidates from Selangor and WP Kuala
Lumpur. Figure 11(c) depicts the pie chart breakdown of the numbers of ineligible candidates that come from
WP Kuala Lumpur, WP Putrajaya, and Negeri Sembilan.

Figure 10. Stacked bar chart of attributes satisfied by eligible applicant(s)

Figure 11. Pie chart of: (a) total applicants based on state distribution in Malaysia, (b) state of four eligible
candidates, and (c) breakdown of the numbers of ineligible candidates

3.4. Usability testing

In calculating the SUS score, the responses provided by participants to the SUS questionnaire were
used. For this study focus, we conducted interviews with three HR expert personnel who were available for the
usability testing due to the time constraints available to complete the study. The questions were divided into

Int J Artif Intell, Vol. 13, No. 4, December 2024: 4334-4343

Int J Artif Intell ISSN: 2252-8938  4341

two types of statements, where questions with odd numbers (1, 3, 5, 7, 9) are phrased positively. For this type,
the respondent’s response score from 1 to 5 was deducted by one. The questions with even numbers
(2, 4, 6, 8, 10) are phrased negatively. Five points will be deducted from the response score for these questions.
The points were tallied together once the adjustments were made for all the questions. The final score was
multiplied by 2.5 to convert the total points into a SUS score on a scale of 0 to 100. This calculation method
allows for a standardized assessment of the system’s perceived usability, with higher SUS scores indicating
better usability. Table 1 shows the findings of the SUS with the raw and final score. The respondents
predominantly selected a scale of 4, representing “agree”, for the positively phrased questions. Conversely, the
respondents mainly selected a scale of 1 for the negatively phrased questions.

Table 1. SUS scores findings

Question No. 1 2 3 4 5 6 7 8 9 10 SUS raw score SUS final score
Expert A 4 1 5 1 4 2 4 1 4 1 35 87.50
Expert B 5 2 4 1 4 1 5 2 5 1 36 90.00
Expert C 4 1 4 1 5 1 5 1 4 1 37 92.50
Average 90.00

The overall percentage for SUS in this study is 90%, which indicates that the system is considered
good. According to Misdan et al. [33], the average SUS score is 68, and scores above 85 are associated with
“excellent”. Scores above average at 71 are presented as “good”, and scores at 51 are considered “ok”. Based
on these criteria, with a SUS score of 90%, the system in this study falls within the excellent rating range,
indicating positive feedback from the participants.

4. CONCLUSION
This study aims to help recruiters identify and select the most suitable candidate who meets the job’s
requirements. Based on the information gathered, we classify and visualize the eligibility rate to find the most
suitable applicants that fit the job offered. The NB model that was applied in this study enables the system to
do the classification tasks. In addition, the diverse visualizations were analyzed from the information gained in
the system application, enabling them to make informed decisions during the hiring process. Clear and intuitive
visualization enables the recruiters to retrieve and summarize information about eligible and ineligible
applicants. The eligibility rate allows recruiters to efficiently handle and gain a comprehensive understanding
of all applicants who meet their criteria in a shorter amount of time. The leveraging power of NB sorts the most
qualified candidates for a position and visualization techniques based on the seven attributes for classification:
title, location, education, years of professional experience, skills, languages, and certificates. For further study,
we recommend adding more functions for contacting the shortlisted applicants from the system and adding
more job fields.

ACKNOWLEDGEMENTS
This research was funded by a grant from Universiti Teknologi MARA Cawangan Melaka (TEJA
Grant 2023 GDT 2023/1-14).

REFERENCES
[1] N. S. Gill, “Recruitment and selection procedures in human resource management,” International Journal of Computer Science and
Mobile Computing, vol. 10, no. 2, pp. 45–49, 2021, doi: 10.47760/ijcsmc.2021.v10i02.006.
[2] P. A. Hamza et al., “Recruitment and selection: The relationship between recruitment and selection with organizational
performance,” International Journal of Engineering, Business and Management, vol. 5, no. 3, 2021, doi: 10.22161/ijebm.5.3.1.
[3] S. I. Abbas, M. H. Shah, and Y. H. Othman, “Critical review of recruitment and selection methods: Understanding the current
practices,” Annals of Contemporary Developments in Management & HR, vol. 3, no. 3, 2021, doi: 10.33166/acdmhr.2021.03.005.
[4] D. Lawong, G. R. Ferris, W. Hochwarter, and L. Maher, “Recruiter political skill and organization reputation effects on job applicant
attraction in the recruitment process: A multi-study investigation,” Career Development International, vol. 24, no. 4, 2019, doi:
10.1108/CDI-01-2019-0007.
[5] G. Dash and D. Chakraborty, “Digital transformation of marketing strategies during a pandemic: Evidence from an emerging
economy during covid-19,” Sustainability, vol. 13, no. 12, 2021, doi: 10.3390/su13126735.
[6] M. S. Hosain and P. Liu, “Linked in for searching better job opportunity: Passive jobseekers’ perceived experience,” Qualitative
Report, vol. 25, no. 10, 2020, doi: 10.46743/2160-3715/2020.4449.
[7] R. Thakkar, “Top 100 hiring statistics for 2022,” LinkedIn, 2022. Accessed: Nov. 24, 2022. [Online]. Available:
https://www.linkedin.com/pulse/top-100-hiring-statistics-2022-rinku-thakkar/
[8] B. T. Janigová, “The genre analysis of job adverts posted on linkedIn,” M.Sc. Thesis, Department of English and American Studies,
Masaryk University, Brno, Czech Republic, 2023.
[9] S. L. -Carril, C. Anagnostopoulos, and P. Parganas, “Social media in sport management education: Introducing LinkedIn,” Journal

Eligibility rate of applicant’s LinkedIn account: a naïve bayes… (Khyrina Airin Fariza Abu Samah)
4342  ISSN: 2252-8938

of Hospitality, Leisure, Sport and Tourism Education, vol. 27, 2020, doi: 10.1016/j.jhlste.2020.100262.
[10] K. S. Wei, “LinkedIn in Malaysia: a comprehensive overview of the growing professional network in 2023,” LinkedIn. 2023.
Accessed: May 04, 2023. [Online]. Available: https://www.linkedin.com/pulse/linkedin-malaysia-comprehensive-overview-
growing-network-shoo/
[11] G. D. Marin and C. Nilă, “Branding in social media using LinkedIn in personal brand communication: A study on
communications/marketing and recruitment/human resources specialists perception,” Social Sciences & Humanities Open, vol. 4,
no. 1, pp. 1–8, 2021, doi: 10.1016/j.ssaho.2021.100174.
[12] B. Hmoud and V. Laszlo, “Will artificial intelligence take over human resources recruitment and selection?,” Network Intelligence
Studies, vol. 7, no. 13, pp. 31–30, 2019.
[13] D. P. Sivanandam and M. P. Mudaliar, “A study on scientific screening process in a recruitment consultancy firm,” Journal of
Contemporary Issues in Business and Government, vol. 26, no. 2, 2021, doi: 10.47750/cibg.2020.26.02.053.
[14] E. Fisher, R. S. Thomas, M. K. Higgins, C. J. Williams, I. Choi, and L. A. McCauley, “Finding the right candidate: Developing
hiring guidelines for screening applicants for clinical research coordinator positions,” Journal of Clinical and Translational Science,
vol. 6, no. 1, 2022, doi: 10.1017/cts.2021.853.
[15] I. Nikolaou, “What is the role of technology in recruitment and selection?,” Spanish Journal of Psychology, vol. 24, 2021, doi:
10.1017/SJP.2021.6.
[16] J. L. R. -Sánchez, T. G. -Torres, A. M. -Navarro, and R. G. -Losada, “Investing time and resources for work–life balance: the effect
on talent retention,” International Journal of Environmental Research and Public Health, vol. 17, no. 6, 2020, doi:
10.3390/ijerph17061920.
[17] H. S. -Szczapa, “Recruitment of employees—assumptions of the risk model,” Risks, vol. 9, no. 3, 2021, doi: 10.3390/risks9030055.
[18] G. Erdoğan, “The ethical shortlisting problem,” Computers and Operations Research, vol. 138, 2022, doi:
10.1016/j.cor.2021.105593.
[19] J. H. Hardy, K. S. Tey, W. C. -Lai, R. F. Martell, A. Olstad, and E. L. Uhlmann, “Bias in context: Small biases in hiring evaluations
have big consequences,” Journal of Management, vol. 48, no. 3, 2022, doi: 10.1177/0149206320982654.
[20] K. A. F. A. Samah, N. S. D. Wirakarnain, R. Hamzah, N. A. Moketar, L. S. Riza, and Z. Othman, “A linear regression approach to
predicting salaries with visualizations of job vacancies: a case study of Jobstreet Malaysia,” IAES International Journal of Artificial
Intelligence, vol. 11, no. 3, pp. 1130–1142, 2022, doi: 10.11591/ijai.v11.i3.pp1130-1142.
[21] M. G. Robinson, “Skills and qualifications for the special library environment in Jamaica: a job advertisement analysis,” Library
Management, vol. 42, no. 1–2, 2021, doi: 10.1108/LM-07-2020-0109.
[22] M. Halinski and J. A. Harrison, “The job resources-engagement relationship: the role of location,” International Journal of Public
Sector Management, vol. 33, no. 6–7, 2020, doi: 10.1108/IJPSM-12-2019-0303.
[23] L. Ronda, C. Abril, and C. Valor, “Job choice decisions: understanding the role of nonnegotiable attributes and trade-offs in effective
segmentation,” Management Decision, vol. 59, no. 6, pp. 1546–1561, 2020, doi: 10.1108/MD-10-2019-1472.
[24] M. Izvercian, S. Potra, and L. Ivascu, “Job satisfaction variables: A grounded theory approach,” Procedia - Social and Behavioral
Sciences, vol. 221, 2016, doi: 10.1016/j.sbspro.2016.05.093.
[25] M. Kaya and T. Bogers, “Effectiveness of job title based embeddings on resume to job ad recommendation,” in 2021 Workshop on
Recommender Systems for Human Resources, RECSYS IN HR 2021, Amsterdam, Netherlands, 2021, pp. 1–7.
[26] N. Chaiyama and N. Kaewpila, “The development of life and career skills in 21st century test for undergraduate students,” European
Journal of Educational Research, vol. 11, no. 1, 2022, doi: 10.12973/eu-jer.11.1.51.
[27] A. Taha et al., “Robotic colorectal surgery: quality assessment of patient information available on the internet using webscraping,”
Computer Assisted Surgery, vol. 28, no. 1, 2023, doi: 10.1080/24699322.2023.2187275.
[28] M. M. Öztürk, “Cosine similarity-based cross-project defect prediction,” Bilişim Teknolojileri Dergisi, vol. 12, no. 3, pp. 159–167,
2019, doi: 10.17671/gazibtd.453436.
[29] L. K. Foo, S. L. Chua, and N. Ibrahim, “Attribute weighted naïve bayes classifier,” Computers, Materials and Continua, vol. 71,
no. 1, 2022, doi: 10.32604/cmc.2022.022011.
[30] K. A. F. A. Samah, N. M. N. Azharludin, L. S. Riza, M. N. H. H. Jono, and N. A. Moketar, “Classification and visualization: Twitter
sentiment analysis of Malaysia’s private hospitals,” IAES International Journal of Artificial Intelligence (IJ-AI), vol. 12, no. 4, pp.
1793–1802, 2023, doi: 10.11591/ijai.v12.i4.pp1793-1802.
[31] P. N. Mwaro, D. K. Ogada, and P. W. Cheruiyot, “Applicability of Naïve Bayes model for automatic resume classification,”
International Journal of Computer Applications Technology and Research, vol. 9, no. 9, pp. 257–264, 2020, doi:
10.7753/ijcatr0909.1002.
[32] Y. Liu and A. I. Abeyratne, Practical applications of Bayesian reliability. Hoboken, United States of America: John Wiley & Sons,
2019.
[33] K. A. F. A. Samah, N. F. A. Misdan, M. N. H. H. Jono, and L. S. Riza, “The best Malaysian airline companies visualization through
bilingual twitter sentiment analysis: A machine learning classification,” JOIV: International Journal on Informatics Visualization,
vol. 6, no. 1, pp. 130–137, 2022, doi: 10.30630/joiv.6.1.879.

BIOGRAPHIES OF AUTHORS

Khyrina Airin Fariza Abu Samah is a senior lecturer at the College of Computing,
Informatics and Mathematics in Universiti Teknologi MARA (UiTM) Melaka Branch, Jasin
Campus. Before joining UiTM, she had 13 years of working experience in the semiconductor
industry. She has a Diploma, Bachelor’s Degree and Master’s Degree in Computer Science and
Ph.D. in Information Technology. Her research interests are in artificial intelligence, algorithm
analysis, machine learning, data science, optimization, and evacuation algorithms. She can be
contacted at email: [email protected].

Int J Artif Intell, Vol. 13, No. 4, December 2024: 4334-4343

Int J Artif Intell ISSN: 2252-8938  4343

Nurul Atirah Ahmad is a Bachelor’s student in Computer Science currently

undergoing her internship programme. Her research interests are data mining, machine learning
algorithms, data analysis and visualization. She is actively involved in the research group at
UiTM. Currently, she is pursuing an internship at Amast Sdn Bhd, Selangor, Malaysia. She can
be contacted at email: [email protected].

Anis Amilah is a lecturer at the College of Computing, Informatics and Mathematics

in Universiti Teknologi MARA (UiTM) Melaka Branch, Jasin Campus. She has a Master of
Sciences degree from UiTM Shah Alam. With a solid educational background, she brings
expertise and knowledge to her role, contributing to the academic environment. Her research
interests revolve around artificial intelligence and emergence algorithms. Her focus on these
cutting-edge fields underscores a commitment to exploring innovative solutions and
understanding the dynamic interplay between intelligence and emergent phenomena. She can be
contacted at email: [email protected].

Hana Fakhira Almarzuki is a lecturer at the College of Computing, Informatics

and Mathematics at Universiti Teknologi MARA (UiTM) Shah Alam, Selangor. She is currently
a Ph.D. candidate in the College of Computing, Informatics and Mathematics, Universiti
Teknologi MARA, Melaka, Jasin Campus, Malaysia. Her research interests are intelligent
tutoring systems, artificial intelligence, and machine learning. She can be contacted at email:
[email protected].

Zuhri Arafah Zulkifli is lecturer at College of Computing, Informatics and

Mathematics in Universiti Teknologi MARA (UiTM) Melaka Branch, Jasin Campus. Before
joining UiTM, she has been working with HSBC GLC Malaysia. She has a Diploma in Computer
Sciences, a Degree in Intelligent Systems from UiTM and a Master of Management Information
Systems from UKM. She has a passionate interest in educational technology, knowledge
management, intelligent systems, and big data. She can be contacted at email:
[email protected].

Lala Septem Riza received Ph.D. in Computer Science from Universidad de

Granada, Spain, in 2015. He works in the Department of Computer Science Education,
Universitas Pendidikan Indonesia, Indonesia. He teaches machine learning, big data platforms,
and statistical data science. His research interests are in machine learning, data science, and
education. He can be contacted at [email protected].

Amir Haikal Abdul Halim received a Master’s Degree in Computer Science from
Universiti Teknologi MARA. He is currently a research assistant and intends to be a Ph.D.
candidate in the College of Computing, Informatics, and Mathematics, Universiti Teknologi
MARA, Melaka, Jasin Campus, Malaysia. He has a Diploma and a Degree in Computer Science.
His research interests are evacuation algorithms, algorithm analysis, artificial intelligence, and
machine learning. He can be contacted at email: [email protected].

Eligibility rate of applicant’s LinkedIn account: a naïve bayes… (Khyrina Airin Fariza Abu Samah)