F1000Research F1000Research 2046-1402 F1000 Research Limited London, UK 10.12688/f1000research.172383.2 Research Article Articles Application of K-Means Clustering for Job Applicant Analysis in Construction Firms Using R

[version 2; peer review: 4 approved with reservations]

 Jaya Daniel Jesayanto Conceptualization Data Curation Formal Analysis Funding Acquisition Investigation Methodology Project Administration Resources Software Supervision Validation Visualization Writing – Original Draft Preparation Writing – Review & Editing https://orcid.org/0000-0003-1940-6302 a 1 2 Ramdhani Wahyu Muhammad Formal Analysis Funding Acquisition Methodology Resources Validation Writing – Review & Editing https://orcid.org/0009-0000-3816-6100 3 Wati Endang Data Curation Funding Acquisition Software Validation Writing – Review & Editing https://orcid.org/0009-0008-4264-7984 3 Nandes Yogi Novario Funding Acquisition Investigation Writing – Review & Editing https://orcid.org/0009-0000-4028-3672 4 Nadhiroh Ilma Zahriyatun Funding Acquisition Resources Writing – Review & Editing https://orcid.org/0009-0002-0352-2691 4 Pade Reza Bakhrun Fidianto Funding Acquisition Software Writing – Review & Editing https://orcid.org/0009-0009-6281-8418 3 Technology and Vocational Education and Training, Universitas Negeri Yogyakarta, Yogyakarta, Special Region of Yogyakarta, 55282, Indonesia Building Engineering Education, Universitas Negeri Jakarta, East Jakarta, Special Capital Region of Jakarta, Indonesia Educational Research and Evaluation, Universitas Negeri Yogyakarta, Yogyakarta, Special Region of Yogyakarta, 55282, Indonesia English Language Education, Universitas Negeri Yogyakarta, Yogyakarta, Special Region of Yogyakarta, 55282, Indonesia danieljesayanto.2023@student.uny.ac.id

No competing interests were disclosed.

 12 3 2026 2025 14 1388 9 2 2026 Copyright: © 2026 Jaya DJ et al. 2026 This is an open access article distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

This study applies K-Means clustering to segment job applicant test data from a construction consulting firm to support data-driven screening decisions. From 161 applicants, 30 candidates who met the document-screening requirements were invited for in-person testing and included in the analysis. Three assessment variables were used: AutoCAD drafting skills, planning and supervision report-writing skills, and adaptability. Using R, K-Means clustering was performed to partition candidates into three groups based on multivariate similarity patterns, and the resulting group structure was visualized using 2D and 3D scatter plots. The clustering output revealed distinct competency profiles: one group characterized by generally lower scores across the three variables, a second group with moderate and mixed scores, and a third group with consistently higher scores. Internal validity indices suggested modest separation (mean silhouette = 0.16; Davies–Bouldin Index = 2.05), consistent with exploratory clustering on a small pre-screened sample. These patterns provide a structured interpretation of applicant diversity and can inform practical recruitment actions such as prioritizing candidates for interviews, identifying borderline profiles for additional evaluation, and designing targeted upskilling recommendations for specific competency gaps. Overall, this study demonstrates how unsupervised clustering of routine recruitment test results can enhance transparency and consistency in early-stage applicant evaluation within construction-sector hiring.

 K-Means Clustering; data-driven recruitment; workforce selection; cluster visualization; construction competencies Lembaga Pengelola Dana Pendidikan 202407112005435 202406111204346 202312211239164 202406111202933 202404111201727 202407111205431 This research was funded by scholarships awarded by the Indonesian Endowment Fund for Education (LPDP), with Grant Numbers 202312211239164 (Daniel Jesayanto Jaya), 202407112005435 (Wahyu Muhammad Ramdhani), 202406111202933 (Endang Wati), 202406111204346 (Yogi Novario Nandes), 202407111205431 (Ilma Zahriyatun Nadhiroh), and 202404111201727 (Reza Bakhrun Fidianto Pade). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Amendments from Version 1

This revised version addresses the reviewers’ methodological, transparency, and interpretive concerns by strengthening analytical rigor, clarifying the study scope, and situating the findings within current discussions on data-driven recruitment and algorithmic governance. The manuscript now provides a clear justification of participant selection, explicitly explaining that clustering was conducted on 30 applicants who passed document screening from an initial pool of 161 applicants. The screening criteria are described in detail, and the analysis is framed as an exploratory examination of a pre-screened assessment sample rather than a predictive or comprehensive recruitment optimization model. Methodological revisions include clearer documentation of data preprocessing, confirmation that all variables were measured on a uniform 0–100 scale, and explicit discussion of centroid initialization and algorithm sensitivity. To address robustness concerns, the clustering analysis was replicated using multiple random initializations in R (nstart), and solution stability was assessed. Internal cluster validity metrics (silhouette coefficient and Davies–Bouldin Index) are now reported as descriptive diagnostics, reinforcing the exploratory nature of the findings. A brief outlier and sensitivity check has also been added, indicating that the cluster structure is not driven by a single extreme observation. To improve readability, detailed centroid iteration and distance calculation tables have been moved to extended data, while the main text focuses on final clustering outcomes and their interpretation. The Results and Discussion sections have been reorganized to emphasize competency profiles, visual interpretation using 2D, 3D, and hierarchical plots, and cautious analytical insights. The Introduction and Discussion have been strengthened by incorporating international perspectives on AI-assisted recruitment, transparency, and fairness, including references to the NIST AI Risk Management Framework, U.S. EEOC guidance on adverse impact under Title VII, and the European Union Artificial Intelligence Act. Data availability and ethical statements have been updated, and extended data have been deposited in Zenodo under a new DOI.

 1. Introduction 1.1 Research background

In the modern workplace, workforce selection is a critical component of human resource development, particularly in sectors that require a combination of technical expertise and adaptive capability. Career development and career transformation are influenced not only by formal qualifications but also by individuals’ ability to adapt to changing work environments and collaborate effectively with diverse stakeholders. Data-driven approaches to workforce analysis have therefore gained attention as tools to support more structured and transparent evaluation processes ( Pala, 2021).

Recruitment involves more than sourcing candidates; it requires systematic decision-making informed by job analysis, organizational needs, and available labor characteristics ( Widodo, 2018). Job analysis plays a central role in defining task requirements, competency expectations, and qualification standards, thereby helping organizations align applicants with role-specific demands. From the applicant’s perspective, successful job search outcomes depend on understanding personal competencies, evaluating labor market opportunities, and developing skills that match employer expectations ( London, 1973).

In the construction sector, technical competencies such as AutoCAD drafting, the ability to prepare planning and supervision reports, and adaptability to dynamic project environments are particularly valued ( Gangl, 2003). These competencies are increasingly important in large-scale infrastructure development contexts. In Indonesia, national strategic projects such as the Nusantara Capital City (Ibu Kota Nusantara, IKN) development have intensified demand for construction personnel with both technical proficiency and social adaptability ( Irmawan et al., 2023; Supriyanti et al., 2023). Managing and interpreting recruitment assessment data in such contexts presents practical challenges, especially when organizations must evaluate multiple competency dimensions simultaneously.

Cluster analysis offers a data-driven approach to explore patterns within applicant assessment data by grouping individuals with similar characteristics. Clustering techniques partition data into internally homogeneous and externally heterogeneous groups, thereby supporting structured interpretation of complex multivariate information ( Jain et al., 1999). Among these techniques, K-Means clustering is widely used due to its computational simplicity and interpretability, making it suitable for exploratory analysis of recruitment-related datasets. In recruitment contexts, clustering can be applied to post-screening assessment data to identify competency profiles rather than to make automated hiring decisions.

Beyond operational efficiency, the use of data-driven tools in recruitment raises broader issues of transparency, governance, and fairness in algorithm-assisted selection. International guidance emphasizes that AI-enabled assessment should be accompanied by risk management, documentation, and ongoing monitoring of unintended impacts ( NIST, 2023). In addition, U.S. Equal Employment Opportunity Commission (EEOC) guidance highlights that employers should assess whether algorithmic or AI-based selection procedures produce adverse impact under Title VII and aligns such assessment with the Uniform Guidelines on Employee Selection Procedures ( EEOC, 2023). Similarly, the European Union Artificial Intelligence Act classifies certain AI systems used in employment-related contexts as high-risk, reinforcing expectations for accountability and safeguards when analytics influence employment decisions ( European Union, 2024). Accordingly, this study positions K-Means clustering as an exploratory decision-support technique rather than an automated hiring system; cluster labels are interpreted cautiously as descriptive competency profiles and are intended to complement human review rather than replace managerial judgment.

This study applies K-Means clustering to recruitment test data from a construction consulting firm, focusing on candidates who passed document screening and completed in-person assessments. Using three core variables—AutoCAD drafting skills, planning and supervision report-writing skills, and adaptability—the study demonstrates how unsupervised clustering can support exploratory analysis of applicant competency profiles within a real organizational context.

 1.2 Literature review

Clustering is an unsupervised analytical technique used to group objects into clusters based on attribute similarity, such that objects within the same cluster exhibit higher similarity than those in other clusters ( Jain et al., 1999). By minimizing within-cluster variation and maximizing between-cluster differences, clustering supports pattern discovery and interpretation in complex datasets ( Manikandan et al., 2018; Darmi & Setiawan, 2016). For organizational and workforce analytics, clustering provides a data-driven means of understanding heterogeneity among individuals without requiring predefined class labels.

Among various clustering approaches, K-Means clustering is one of the most widely applied methods due to its simplicity, efficiency, and interpretability. K-Means partitions data into k clusters by iteratively assigning observations to the nearest centroid and updating centroid positions until convergence is achieved ( Jain et al., 1999). Because of its relatively low computational cost, K-Means is suitable for applied settings where rapid analysis and transparent interpretation are required ( Fadhli, 2017).

Previous studies demonstrate applicability across domains. In educational research, K-Means has been used to analyze student preferences and learning achievement patterns ( Firza & Sarjono, 2020). In organizational contexts, it has been applied to group employees based on discipline and performance indicators to support human resource decision-making ( Agustina & Prihandoko, 2018). Comparative studies suggest that while alternatives such as Fuzzy C-Means may offer advantages in some conditions, K-Means remains computationally efficient and practical for many real-world applications ( Wiharto & Suryani, 2020).

1.2.1 K-Means algorithm

K-Means is a partition-based clustering algorithm that divides data into a predefined number of clusters by minimizing the average distance between data points and their respective cluster centroids ( Widiyaningtyas et al., 2017). The algorithm operates iteratively, beginning with the selection of initial centroid values and proceeding through repeated reassignment of data points based on distance calculations until cluster membership stabilizes ( Purba et al., 2018). Prior work emphasizes that K-Means can be sensitive to initialization and the scale of input variables, highlighting the need for transparent methodological choices in applied studies ( Jain et al., 1999).

1.2.2 Worker recruitment

Recruitment is a strategic organizational process aimed at attracting and selecting individuals whose competencies align with job requirements and organizational objectives. Job analysis plays a critical role in defining tasks, responsibilities, and qualification standards, thereby guiding recruitment and selection decisions ( Widodo, 2018). In the construction sector, recruitment emphasizes a combination of technical competencies—such as drafting and report preparation—and adaptive capabilities, reflecting the dynamic and collaborative nature of construction projects ( Gangl, 2003). The job search process seeks to match job seekers with appropriate opportunities and can be supported through technology-enabled and data-driven methods ( Green et al., 2011). Given the multidimensionality of applicant data, clustering methods such as K-Means offer a way to organize assessment results into interpretable competency profiles that can support early-stage evaluation ( Jain et al., 1999).

 2. Methods 2.1 Research Design

This study employed a quantitative, exploratory research design using unsupervised clustering to analyze recruitment assessment data from a construction consulting firm. The primary objective was to explore competency-based grouping patterns among job applicants using K-Means clustering as a decision-support tool, rather than to predict hiring outcomes or evaluate post-employment performance.

Workflow research diagram. 2.2 Data source and participant selection

The data were obtained from CV Ardantama Putra Perkasa as part of its internal recruitment process. Although the vacancy was advertised through JobStreet Indonesia, all data analyzed in this study originated exclusively from the company’s internal screening and testing procedures.

A total of 161 applicants applied for the position. Applicants were shortlisted through the company’s standard document-screening procedure conducted by the HR team and the hiring unit. Screening focused on administrative completeness and role relevance, including:

(i) completeness of required documents;

(ii) educational background and relevance to construction consulting work;

(iii) evidence of relevant technical exposure (e.g., drafting/reporting-related tasks or portfolio where available); and

(iv) basic eligibility criteria specified in the vacancy announcement.

From this screening stage, 30 candidates who met the minimum requirements were invited for in-person testing. Only these 30 candidates were included in the clustering analysis because complete assessment scores were available for all three variables. This design improves internal consistency of the tested dataset but limits generalizability to the full applicant pool.

 2.3 Assessment variables

Candidates were evaluated using three competency indicators relevant to construction consulting roles:

AutoCAD drafting skills;

planning and supervision report-writing skills;

adaptability.

Each variable was assessed on a numerical scale from 0 to 100, with higher scores indicating stronger performance.

 2.4 Data preprocessing, outlier, and sensitivity checks

The dataset was reviewed for completeness and consistency. All 30 candidates had complete scores across the three assessment variables; therefore, no records were excluded at this stage. Because all variables were measured using the same scale (0–100), the analysis used raw scores without additional normalization to preserve the meaning of the original assessment scores.

A basic outlier and sensitivity check was conducted by examining distances to cluster centroids and visually inspecting the 3D scatter plot. A leave-one-out sensitivity test removing the most distant observation did not materially change the overall three-cluster interpretation; validity metrics changed only slightly (mean silhouette increased from 0.16 to approximately 0.18; DBI remained approximately 2.0). This suggests the reported structure is not driven by a single extreme case.

 2.5 Clustering procedure

K-Means clustering was applied to group candidates based on similarity across the three assessment variables. The number of clusters was set to k = 3, reflecting the company’s practical need to differentiate candidates into three evaluative groups for recruitment support.

Initial centroid values were specified as starting points based on preliminary inspection of score distributions during exploratory analysis. These initial values were used to initiate iteration rather than to impose predetermined outcome categories. Euclidean distance was used to assign candidates to the nearest centroid, after which centroid positions were updated as the mean of cluster members. The algorithm iterated until cluster assignments stabilized.

The clustering workflow was implemented using a combination of spreadsheet-based calculations (for transparency of manual steps) and the R programming language (for reproducibility, validity checks, and visualization). Intermediate iteration tables are provided as extended data.

2.5.1 Initialization and stability checks

Because K-Means can be sensitive to initialization, the analysis was repeated in R using the built-in kmeans() function with multiple random initializations (e.g., nstart = 50). Solution stability was assessed by comparing convergence outcomes (within-cluster sum of squares) and checking consistency of cluster memberships across repeated initializations. This step ensured that the reported three-cluster structure was not an artifact of a single starting configuration. Minor membership differences across runs occurred for borderline profiles, which is plausible in small samples with overlapping competency distributions.

 2.6 Visualization and interpretation

Clustering results were visualized using two-dimensional and three-dimensional scatter plots. Two-dimensional plots illustrated relationships between AutoCAD drafting skills and planning/supervision report-writing skills, while three-dimensional plots incorporated adaptability as a third axis.

Clusters were subsequently labeled as “Rejected,” “Under Consideration,” and “Accepted” based on their relative position in the multivariate competency space. These labels represent analytical interpretations of score patterns and do not constitute formal hiring decisions made by the company.

 2.7 Scope and methodological limitations

This study focuses on exploratory grouping of recruitment assessment data from a pre-screened subset of applicants. The clustering results were not validated against final hiring decisions or post-employment performance outcomes. Accordingly, findings should be interpreted as structured analytical support rather than definitive evidence of selection effectiveness.

 2.8 Cluster validity assessment

To provide quantitative support for the cluster structure, internal validity indices were calculated. The silhouette coefficient was computed using Euclidean distances to estimate how well each candidate matched its assigned cluster relative to other clusters. The Davies–Bouldin Index (DBI) was calculated to evaluate average cluster similarity based on within-cluster dispersion relative to between-cluster centroid distances. These indices were interpreted as descriptive diagnostics of separation quality rather than evidence of predictive utility.

 3. Results and discussion 3.1 Applicant characteristics

This study analysed recruitment assessment records from CV Ardantama Putra Perkasa, obtained from the company’s internal testing and selection process. A total of 161 applicants submitted applications, of whom 30 candidates meeting minimum screening criteria were invited for in-person testing. Each candidate was assessed on three indicators measured on a 0–100 scale: AutoCAD drafting skills (X), planning and supervision report-writing skills (Y), and adaptability (Z). Candidate characteristics and scores are summarised in Table 1.

Applicant demographic data.
Respondent code Gender AutoCAD drawing skills (X) Ability to prepare planning and monitoring reports (Y) Adaptability (Z)
Resp1 Female 92 75 68
Resp2 Male 68 65 66
Resp3 Male 73 86 87
Resp4 Male 69 74 73
Resp5 Male 78 72 91
Resp6 Female 84 90 92
Resp7 Male 69 76 87
Resp8 Female 95 73 76
Resp9 Female 90 80 85
Resp10 Male 68 82 68
Resp11 Male 63 75 71
Resp12 Male 75 93 77
Resp13 Female 62 72 68
Resp14 Male 90 61 72
Resp15 Female 84 63 90
Resp16 Female 94 70 89
Resp17 Female 73 87 80
Resp18 Female 71 73 95
Resp19 Female 93 62 70
Resp20 Male 90 68 89
Resp21 Female 87 94 87
Resp22 Male 60 90 64
Resp23 Female 65 64 93
Resp24 Male 69 84 75
Resp25 Male 66 63 72
Resp26 Male 95 85 93
Resp27 Male 75 80 83
Resp28 Male 92 85 93
Resp29 Male 71 71 85
Resp30 Male 92 61 88

Overall, the score distribution shows meaningful heterogeneity across candidates—particularly in adaptability and planning/supervision report-writing—indicating variation in both technical and interpersonal readiness. This variability provides a suitable basis for exploratory clustering analysis.

 K-Means clustering results

Using K-Means clustering with k = 3, the 30 assessed candidates were grouped into three distinct clusters based on similarity across AutoCAD drafting skills, planning and supervision report-writing skills, and adaptability. The final cluster assignments are summarized in Table 2. These clusters represent analytical competency profiles derived from multivariate similarity patterns rather than formal hiring decisions determined by company policy.

Final clustering results.
Respondent data Rejected Under consideration Accepted Clustering
C1(x 1,y 1,z 1) C2(x 2,y 2,z 2) C3(x 3,y 3,z 3)
Name MA LPP KA 67,30 78,50 71,40 73,25 73,13 88,88 91,17 75,33 83,50
Resp2 68 65 66 14,56 24,84 30,82 Cluster 1 (Rejected)
Resp4 69 74 73 5,07 16,46 24,56
Resp10 68 82 68 4,93 23,28 28,66
Resp11 63 75 71 5,56 20,69 30,82
Resp13 62 72 68 9,05 23,74 33,20
Resp22 60 90 64 15,50 32,85 39,58
Resp24 69 84 75 6,79 18,13 25,27
Resp25 66 63 72 15,57 20,97 30,29
Resp17 73 87 80 13,37 16,47 21,87
Resp12 75 93 77 17,35 23,22 24,81
Resp7 69 76 87 15,89 5,46 22,45 Cluster 2 (Under Consideration)
Resp23 65 64 93 26,12 12,97 30,06
Resp29 71 71 85 15,97 4,96 20,68
Resp5 78 72 91 23,26 5,32 15,52
Resp18 71 73 95 24,51 6,53 23,33
Resp3 73 86 87 18,22 13,01 21,36
Resp15 84 63 90 29,41 14,81 15,68
Resp27 75 80 83 14,00 9,21 16,83
Resp14 90 61 72 28,67 26,69 18,41 Cluster 3 (Accepted)
Resp19 93 62 70 30,57 29,50 19,06
Resp1 92 75 68 25,18 28,12 15,53
Resp6 84 90 92 28,91 20,25 18,40
Resp8 95 73 76 28,61 25,28 8,74
Resp9 90 80 85 26,50 18,52 5,04
Resp16 94 70 89 33,09 20,98 8,17
Resp20 90 68 89 30,58 17,52 9,24
Resp21 87 94 87 29,52 25,07 19,44
Resp26 95 85 93 35,72 25,12 14,09
Resp28 92 85 93 33,45 22,57 13,58
Resp30 92 61 88 34,52 22,35 15,05

The first cluster is characterized by relatively lower combined scores across the three assessed competencies. The second cluster consists of candidates with moderate and mixed competency scores, reflecting intermediate profiles that may warrant further evaluation. The third cluster comprises candidates with consistently higher scores across technical and adaptive dimensions, indicating stronger and more balanced competency profiles.

The clustering process involved iterative centroid updates until cluster memberships stabilized. To maintain readability, detailed iteration tables are provided as extended data, while the main text focuses on the stabilized results and their interpretation. Re-running clustering with multiple random initializations in R produced highly similar solutions, suggesting the three-cluster structure was not dependent on a single manual initialization. Minor membership differences across runs occurred for borderline profiles, which is expected in small samples with partially overlapping competency distributions.

The final clustering output generated from the R environment, including cluster labels and competency scores for each applicant, is presented in Table 3.

R-generated data table.
No AutoCAD_Drafting Planning_Supervision_ Reports Adaptability Cluster Category
1 68 65 66 1 Rejected
2 69 74 73 1 Rejected
3 68 82 68 1 Rejected
4 63 75 71 1 Rejected
5 62 72 68 1 Rejected
6 60 90 64 1 Rejected
7 69 84 75 2 Under Consideration
8 66 63 72 1 Rejected
9 73 87 80 2 Under Consideration
10 75 93 77 2 Under Consideration
11 69 76 87 3 Accepted
12 65 64 93 1 Rejected
13 71 71 85 2 Under Consideration
14 78 72 91 3 Accepted
15 71 73 95 2 Under Consideration
16 73 86 87 2 Under Consideration
17 84 63 90 3 Accepted
18 75 80 83 2 Under Consideration
19 90 61 72 2 Under Consideration
20 93 62 70 2 Under Consideration
21 92 75 68 2 Under Consideration
22 84 90 92 3 Accepted
23 95 73 76 3 Accepted
24 90 80 85 3 Accepted
25 94 70 89 3 Accepted
26 90 68 94 3 Accepted
27 87 84 87 3 Accepted
28 95 85 93 3 Accepted
29 92 85 93 3 Accepted
30 92 61 88 2 Under Consideration

3.2.1 Cluster validity metrics

Internal validation indicated modest cluster separation. The mean silhouette coefficient was 0.16, suggesting partial overlap among competency profiles, which is plausible given the small pre-screened sample. The Davies–Bouldin Index was 2.05, indicating moderate distinctiveness among the three clusters. These values support interpreting the clusters as exploratory competency groupings rather than sharply separated classes.

 3.3 Visualization of cluster structure

To support interpretation, two-dimensional and three-dimensional visualizations were generated. Figure 2 presents a 2D scatter plot based on AutoCAD drafting skills and planning/supervision report-writing skills, showing visible separation between lower, intermediate, and higher competency profiles along key technical dimensions.

K-means clustering visualization in a 2D scatter plot.

Figure 3 extends the visualization into three dimensions by incorporating adaptability as a third axis. The 3D scatter plot reveals clearer spatial separation among clusters, particularly distinguishing candidates who combine strong technical skills with high adaptability from those with lower overall competency scores.

K-means clustering visualization in a 3D scatter plot.

For clarity of interpretation, the clustered dataset sorted by category is provided in Table 4.

Sorted dataset by cluster categories.
No AutoCAD_Drafting Planning_Supervision_ Report Adaptability Cluster Category
1 68 65 66 1 Rejected
2 69 74 73 1 Rejected
3 68 82 68 1 Rejected
4 63 75 71 1 Rejected
5 62 72 68 1 Rejected
6 60 90 64 1 Rejected
7 66 63 72 1 Rejected
8 65 64 93 1 Rejected
9 69 84 75 2 Under Consideration
10 73 87 80 2 Under Consideration
11 75 93 77 2 Under Consideration
12 71 71 85 2 Under Consideration
13 71 73 95 2 Under Consideration
14 73 86 87 2 Under Consideration
15 75 80 83 2 Under Consideration
16 90 61 72 2 Under Consideration
17 93 62 70 2 Under Consideration
18 92 75 68 2 Under Consideration
19 92 61 88 2 Under Consideration
20 69 76 87 3 Accepted
21 78 72 91 3 Accepted
22 84 63 90 3 Accepted
23 84 90 92 3 Accepted
24 95 73 76 3 Accepted
25 90 80 85 3 Accepted
26 94 70 89 3 Accepted
27 90 68 89 3 Accepted
28 87 94 87 3 Accepted
29 95 85 93 3 Accepted
30 92 85 93 3 Accepted

To further examine structural consistency, hierarchical clustering projected onto principal component space is presented in Figure 4. Although hierarchical clustering was not employed as the primary analytical method, the observed grouping patterns broadly align with the K-Means classification, providing additional support for the stability of the three-cluster structure within this dataset.

Hierarchical clustering visualization using PCA-projected dimensions. 3.4 Interpretation and discussion

The clustering results demonstrate that K-Means can be used as an exploratory tool to organize recruitment assessment data into interpretable competency profiles within a construction consulting context. Candidates grouped in the higher-scoring cluster tend to exhibit stronger performance across both technical and adaptive dimensions, consistent with prior research emphasizing the importance of combining technical competence with adaptability in project-based and construction-related work environments ( Gangl, 2003; Brown & Hesketh, 2005).

The intermediate cluster represents candidates with mixed strengths, suggesting development potential rather than clear acceptance or rejection outcomes. This aligns with literature highlighting the role of structured training and targeted skill development in enhancing workforce readiness and career progression ( Rawat et al., 2024). Rather than constituting definitive recruitment decisions, this cluster highlights individuals who may benefit from managerial attention, follow-up assessment, or additional training.

Importantly, the clustering approach does not replace professional judgment in recruitment. Instead, it provides a structured analytical perspective that can support transparency and consistency in early-stage evaluation. This aligns with contemporary views that HR analytics is most effective when it complements human expertise rather than automates decision-making processes ( Akkermans et al., 2024).

From an ethical and governance perspective, the analysis is intended to structure early-stage assessment information rather than to automate acceptance decisions. Guidance on trustworthy AI and employment decision tools emphasizes the need for documentation, monitoring, and attention to bias risks when analytics are used in consequential settings ( NIST, 2023; EEOC, 2023). Accordingly, the cluster labels in this study are treated as descriptive competency profiles and should be used alongside human review, transparent documentation, and periodic evaluation of potential disparate impact.

 3.5 Methodological considerations and limitations

Several limitations should be considered when interpreting these findings. First, the analysis was conducted on a pre-screened subset of candidates who passed document screening and participated in in-person testing; therefore, results may not generalize to the full applicant pool. Second, internal validity indices indicated modest separation, suggesting partially overlapping competency profiles that are plausible in a small filtered sample. Third, the clusters were not externally validated against final hiring decisions, expert HR evaluation, or subsequent job performance outcomes.

Despite these limitations, the results illustrate how clustering can function as a practical decision-support tool in recruitment contexts involving multidimensional competency assessments. Future research could extend this approach by applying clustering to larger and more diverse applicant pools, incorporating additional competency indicators, comparing alternative clustering methods, and validating cluster profiles against post-hire performance indicators.

 4. Conclusions

This study explored the use of K-Means clustering as an exploratory analytical approach for organizing recruitment assessment data in a construction consulting context, based on three competencies: AutoCAD drafting skills, planning and supervision report-writing skills, and adaptability. Using data from a pre-screened group of applicants, the analysis identified three distinct competency profiles reflecting different patterns of technical and adaptive capabilities.

The identified clusters indicate that applicants with stronger and more balanced combinations of technical competence and adaptability tend to form a distinct group, while candidates with mixed or lower competency profiles are grouped separately. These results should be interpreted as analytical groupings based on similarity patterns rather than as definitive hiring decisions or evidence of predictive effectiveness. This interpretation is consistent with conceptual discussions emphasizing the importance of adaptability and skill alignment in contemporary labor markets ( Gangl, 2003; Brown & Hesketh, 2005).

Quantitative diagnostics suggested modest separation (mean silhouette = 0.16; DBI = 2.05), supporting cautious interpretation of the clusters as exploratory profiles in a small screened sample. The use of two-dimensional and three-dimensional visualizations enhanced interpretability by illustrating how multivariate competency combinations differentiate applicant profiles. The observed alignment between K-Means results and supporting hierarchical visualization further suggests structural consistency within the analyzed dataset, although external validation against hiring outcomes or job performance was beyond the scope of this study.

From a practical standpoint, the findings suggest that clustering-based analysis may support early-stage recruitment evaluation by helping organizations structure and interpret multidimensional assessment data in a transparent and systematic manner. More broadly, clustering as a decision-support mechanism can be situated within wider discussions on data-driven analysis as a means of structuring managerial judgment rather than replacing it ( Diván, 2017). When used alongside professional expertise, such approaches align with contemporary perspectives on human resource analytics that emphasize analytical support over automated decision-making ( Akkermans et al., 2024).

In addition, the presence of an intermediate competency cluster highlights applicants who may benefit from further evaluation or targeted skill development initiatives, echoing research on structured training and career development ( Rawat et al., 2024; Donald et al., 2024). While career sustainability and job insecurity were not directly examined, the inclusion of adaptability as a clustering dimension resonates with broader discussions on adaptive capacity in uncertain career contexts ( Van der Heijden et al., 2024).

Overall, this study provides a practical illustration of how unsupervised clustering techniques can be applied to recruitment assessment data in the construction sector. By emphasizing transparency, interpretability, and cautious use of analytics within governance and fairness considerations ( NIST, 2023; EEOC, 2023), the study contributes to ongoing discussions on data-driven decision-support tools for workforce selection and development.

 Ethical approval

Ethical review and approval were not required for this study because the researchers analyzed fully anonymized secondary data that had been lawfully transferred by CV Ardantama Putra Perkasa under a formal Data Usage Agreement (No. 12/X/S-K/APP/2024). According to Indonesian national research ethics regulations (Permenkes RI No. 74/2016, Article 11) and the general principles of the Declaration of Helsinki, research involving secondary anonymized non-clinical data that cannot identify individuals is exempt from institutional ethical review. Therefore, this study qualifies for an ethics exemption.

 Informed consent

Informed consent for data use was not obtained directly by the researchers, as all data were collected by CV Ardantama Putra Perkasa under standard recruitment procedures. The company confirmed, through the Data Usage Agreement (No. 12/X/S-K/APP/2024), that job applicants had authorized the use of their anonymized recruitment test results for evaluation and administrative purposes in accordance with Indonesian data protection regulations (UU ITE and PP 71/2019). Because the researchers received only anonymized secondary data and had no access to identifiable information, this study meets the criteria for consent exemption.

 Clinical trial registration

Not applicable.

 Data availability statement Underlying data

The anonymized job applicant dataset is not publicly available due to confidentiality agreements with CV Ardantama Putra Perkasa. Access may be granted for legitimate academic research upon reasonable request to the corresponding author ( danieljesayanto.2023@student.uny.ac.id), subject to approval by the data owner and compliance with Indonesian data protection regulations (UU ITE and PP 71/2019), including signing a Data Use Agreement and a commitment not to attempt re-identification.

 Extended data

Extended data supporting this study, including R scripts, clustering iteration tables, visualizations, and documentation, are openly available in Zenodo at https://doi.org/10.5281/zenodo.18501546 ( Jaya, 2026) under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.

 Acknowledgement

The authors gratefully acknowledge the financial support provided by the Indonesian Endowment Fund for Education (LPDP) as the official sponsor of the scholarships that supported this publication. The authors also thank CV Ardantama Putra Perkasa for granting formal permission to use anonymized job applicant data for research and academic purposes.

 References Akkermans J Donald WE Jackson D : Are we talking about the same thing? The case for stronger connections between graduate and worker employability research. Career Dev. Int. 2024;29(1):80–92. 10.1108/CDI-08-2023-0278 Agustina N Prihandoko P : Perbandingan algoritma K-Means dengan Fuzzy C-Means untuk clustering tingkat kedisiplinan kinerja karyawan. Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi). 2018;2(3):621–626. 10.29207/resti.v2i3.492 Brown P Hesketh A : The mismanagement of talent: Employability and jobs in the knowledge economy. Ind. Labor Relat. Rev. 2005. 10.2189/asqu.2005.50.2.306 Chen Yu : K-Means clustering. Indiana University;2020. Darmi Y Setiawan A : Penerapan metode clustering K-Means dalam pengelompokan penjualan produk. Jurnal Media Infotama. 2016;12(2):148–157. Diván M : Data-driven decision making. 2017 IEEE International Conference on Technological Innovations in ICT for Agriculture and Rural Development (TIAR). IEEE;2017; pp.50–56. 10.1109/ICTUS.2017.8285973 Donald WE Van der Heijden BIJM Manville G : (Re) Framing sustainable careers: Toward a conceptual model and future research agenda. Career Dev. Int. 2024;29(5):513–526. 10.1108/CDI-02-2024-0073 EEOC : Select Issues: Assessing Adverse Impact in Software, Algorithms, and Artificial Intelligence Used in Employment Selection Procedures Under Title VII of the Civil Rights Act of 1964 (Technical Assistance; EEOC-NVTA-2023-2, Issue Date: 2023-05-18). U.S. Equal Employment Opportunity Commission;2003. https://data.aclum.org/storage/2025/01/EOCC_www_eeoc_gov_laws_guidance_select-issues-assessing-adverse-impact-software-algorithms-and-artificial.pdf El Achmar D Bhagat R : The conceptual relation between human resource management (HRM) and competency mapping. International Journal of Teaching & Education. 2023. European Union Parliament and Council : Regulation (EU) 2024/1689 … (Artificial Intelligence Act).Official Journal of the European Union, OJ L, 2024/1689, 12.7.2024. EUR-Lex.2024. https://data.europa.eu/eli/reg/2024/1689/oj Fadhli M : Manajemen peningkatan mutu pendidikan. Tadbir: Jurnal Studi Manajemen Pendidikan. 2017;1(2):215–240. 10.29240/jsmp.v1i2.295 Firza F Sarjono S : Penerapan algoritma K-Means dalam metode clustering untuk peminatan jurusan bagi siswa Swasta Pelita Raya Kota Jambi. Jurnal Manajemen Sistem Informasi. 2020;5(3):371–382. Gangl M : Labor market structure and re-employment rates: Unemployment dynamics in West Germany and the United States. Research in Social Stratification and Mobility. 2003;20:185–224. 10.1016/S0276-5624(03)20004-4 Gie W Jollyta D : Perbandingan Euclidean dan Manhattan untuk optimasi cluster menggunakan Davies-Bouldin Index: Status COVID-19 wilayah Riau. Prosiding Seminar Nasional Riset Information Science (SENARIS). 2020, July;2:187–191. Green AE Hoyos M Li Y : Job search study: Literature review and analysis of the Labour Force Survey. London: Department for Work and Pensions;2011. Hurbean L Miliaru F Muntean M : The impact of business intelligence and analytics adoption on decision-making effectiveness and managerial work performance. Scientific Annals of Economics and Business. 2023;70:43–54. 10.47743/saeb-2023-0012 Irmawan I Sagharmata FA Ruthriana F : Analisis dampak pembangunan Kota Hutan (Forest City) (Studi kasus: Ibu Kota Nusantara (IKN), Kalimantan). Prosiding Seminar Rekayasa Teknologi (SemResTek). 2023;299–304. Jain AK Murty MN Flynn PJ : Data clustering: A review. ACM Computing Surveys (CSUR). 1999;31(3):264–323. 10.1145/331499.331504 Jaya DJ : Supplementary Materials R1 for “Application of K-Means Clustering for Job Applicant Analysis in Construction Firms Using R”.[Data set]. Zenodo. 2026. 10.5281/zenodo.18501546 Kassambara A : Practical guide to cluster analysis in R. STHDA; 1st ed. 2017. Reference Source London HH : Principles and techniques of vocational guidance. Ohio: Charles E. Merrill Publishing Company;1973. Manikandan S Caroline AL Kanniamma D : The study on clustering analysis in data mining. International Journal of Data Mining Techniques and Applications. 2018;7(1):46–49. National Institute of Standards and Technology (NIST) : Artificial Intelligence Risk Management Framework (AI RMF 1.0). 2023. https://doi.org/10.6028/NIST.AI.100-1 Pala SK : Use and applications of data analytics in human resource management and talent acquisition. International Journal of Enhanced Research in Science, Technology & Engineering. 2021;10:2319–7463. Purba W Tamba S Saragih J : The effect of mining data K-Means clustering toward students profile model drop out potential. IOP Conference Series: Journal of Physics. 2018;1007(1):012046–012049. 10.1088/1742-6596/1007/1/012049 Rawat A Nadavulakere S Isenhour L : Career enhancement strategies, supportive work relationships and subjective career success: The moderating role of family–work conflict. Career Dev. Int. 2024;29(4):421–433. 10.1108/CDI-06-2023-0160 Smith SC Todaro MP : Economic development. Boston: Pearson Education; 12th ed. 2015. Supriyanti SS Kusmayanti JD Paluseri ARA : Pemberdayaan masyarakat sekitar di wilayah Ibu Kota Nusantara. Masyarakat Indonesia. 2023;49(1):93–102. Van der Heijden BIJM Hofer A Semeijn J : “Don’t you worry ’bout a thing” – The moderating role of age in the relationship between qualitative job insecurity and career sustainability. Career Dev. Int. 2024;29(5):527–543. 10.1108/CDI-08-2023-0280 Widiyaningtyas T Prabowo MIW Pratama MAM : Implementation of K-Means clustering to distribution of high school teachers. Proceeding EECSI, Yogyakarta, 19–21 September. 2017, September;49–54. Widodo SE : Manajemen pengembangan sumber daya manusia. Yogyakarta: Pustaka Pelajar;2018. Wiharto W Suryani E : The comparison of clustering algorithms K-Means and Fuzzy C-Means for segmentation retinal blood vessels. Acta Informatica Medica. 2020;28(1):42–46. 10.5455/aim.2020.28.42-47 Zhang M Zhou S Wu Y : Pressure from social media: Influence of social media usage on career exploration. Career Dev. Int. 2024;29(1):93–112. 10.1108/CDI-01-2023-0016 10.5256/f1000research.196340.r468949 Reviewer response for version 2 Pişirgen Ali 1 Referee https://orcid.org/0000-0001-7257-2938 Karamanoğlu Mehmetbey University, Karaman, Turkey

Competing interests: No competing interests were disclosed.

 20 4 2026 Copyright: © 2026 Pişirgen A 2026 This is an open access peer review report distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. recommendation approve-with-reservations

Choosing the number of clusters

There is an inherent tension: K=3 is motivated by desired recruitment categories, but K should be justified from the data structure as well, otherwise the analysis becomes forced classification by clustering.

The manuscript should present cluster-number justification more rigorously (e.g., elbow method plus at least one additional diagnostic such as silhouette/gap or stability), and explain that “three clusters” may be a managerial convenience rather than a natural structure.

Terminology discipline:

Use one consistent naming scheme for variables (e.g., “AutoCAD drafting” vs “AutoCAD drawing”; “planning/supervision report writing”). Inconsistencies make reproducibility and interpretation harder, especially if extended data uses different names. Particularly Table 1 and Table 4. Ensure that variable names match

Reproducibility

Because this is an applied observational dataset analysis (not a clinical trial, systematic review, or animal study), CONSORT/PRISMA/ARRIVE are not directly applicable; however, STROBE-style completeness standards are useful for reporting observational data and analysis decisions. 1

Is the work clearly and accurately presented and does it cite the current literature?

Yes

If applicable, is the statistical analysis and its interpretation appropriate?

Yes

Are all the source data underlying the results available to ensure full reproducibility?

No source data required

Is the study design appropriate and is the work technically sound?

Partly

Are the conclusions drawn adequately supported by the results?

Yes

Are sufficient details of methods and analysis provided to allow replication by others?

Yes

Reviewer Expertise:

My research focuses on data analytics and decision support systems within the field of information systems, with particular emphasis on machine learning, clustering techniques, and data-driven modeling of socio-economic systems such as scientometrics, tourism, marketing.

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.

References : The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: Guidelines for Reporting Observational Studies. PLoS Medicine .2007;4(10) : 10.1371/journal.pmed.0040296 10.1371/journal.pmed.0040296 Ramdhani Wahyu Muhammad Educational Research and Evalu, Universitas Negeri Yogyakarta, Yogyakarta, Special Region of Yogyakarta, Indonesia

Competing interests: we declare that we have no competing interest in this work.

 7 5 2026

Dear Reviewer,

Thank you for your careful and constructive review of our manuscript and for approving the work with reservations. We have revised the manuscript to address each of your comments.

1. Choosing the number of clusters

Comment:

There is an inherent tension: K = 3 is motivated by desired recruitment categories, but K should be justified from the data structure as well. The manuscript should present cluster-number justification more rigorously and explain that three clusters may be a managerial convenience rather than a natural structure.

Response:

We agree with this important point. In the revised manuscript, we now explicitly state that the three-cluster solution was selected partly because it corresponded to the organization’s need for an interpretable decision-support structure, but that it should not be interpreted as proof of three naturally distinct applicant groups. We have reframed the three-cluster solution as a managerially interpretable exploratory grouping rather than a definitive natural classification.

We also expanded the cluster-number justification by discussing internal validity diagnostics, including the elbow method, silhouette coefficient, and Davies–Bouldin Index. The revised manuscript reports a mean silhouette coefficient of 0.16 and a Davies–Bouldin Index of 2.05, which are interpreted cautiously as indicating modest separation rather than strong natural cluster structure.

Changes made:

The Methods and Results sections now include clearer discussion of cluster-number justification, managerial interpretability, and internal validity diagnostics. The Results, Discussion, and Conclusions have also been revised to avoid overstating the strength or naturalness of the three-cluster solution.

2. Terminology discipline

Comment:

Use one consistent naming scheme for variables, especially in Table 1 and Table 4. Inconsistencies make reproducibility and interpretation harder.

Response:

We agree. We have standardized the terminology across the manuscript, tables, figures, and supplementary materials. The revised manuscript consistently uses the following terms: “AutoCAD drafting skills,” “planning/supervision report-writing skills,” and “adaptability.” We also revised the tables and figure labels to align with this terminology. The previous decision-style labels were replaced with neutral interpretive profile labels to avoid implying automated hiring decisions.

Changes made:

Table 1 and the results tables have been revised using consistent terminology. The labels “Rejected,” “Under Consideration,” and “Accepted” were replaced with “Lower competency profile,” “Intermediate/mixed competency profile,” and “Higher competency profile.” Revised figures and supplementary materials were also provided.

3. Reproducibility and reporting completeness

Comment:

Because this is an applied observational dataset analysis, CONSORT/PRISMA/ARRIVE are not directly applicable; however, STROBE-style completeness standards are useful for reporting observational data and analysis decisions.

Response:

We agree. We have revised the manuscript to follow STROBE-style completeness principles where relevant. The revised Methods section now clarifies the source of the data, the applicant screening process, the reason for analysing only the 30 shortlisted candidates, the assessment variables, preprocessing decisions, clustering parameters, and limitations of inference.

Changes made:

We added explicit explanation of the analytical sample, eligibility process, preprocessing decisions, and reporting boundaries. We also cited the STROBE statement and revised the limitations section to clarify the observational and exploratory nature of the study.

4. Use of supplementary materials

Response:

We also updated the supplementary materials to improve consistency and reproducibility. The revised extended data include the final R script, clustering outputs, revised figure files, and supporting documentation. The supplementary terminology was updated to match the revised manuscript.

We appreciate the reviewer’s comments, which helped us improve the methodological discipline, terminology consistency, and interpretation of the clustering results.

 10.5256/f1000research.196340.r467509 Reviewer response for version 2 Kembuan Olivia 1 Referee https://orcid.org/0000-0003-4549-0544 Sangkop Ferdinan 2 Co-referee https://orcid.org/0000-0002-6361-7919 Universitas Negeri Manado, Sulawesi Utara, Indonesia informatics, Universitas Negeri Manado (Ringgold ID: 175496), Tondano, North Sulawes, Indonesia

Competing interests: No competing interests were disclosed.

 23 3 2026 Copyright: © 2026 Kembuan O and Sangkop F 2026 This is an open access peer review report distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. recommendation approve-with-reservations

Is the work clearly and accurately presented and does it cite the current literature?

The manuscript is generally clearly structured and readable.  However, the literature review focuses primarily on clustering algorithms and includes a number of regional or context-specific references. The manuscript would benefit from stronger engagement with more recent international literature on HR analytics, algorithmic decision-making in recruitment, and AI-assisted hiring systems. Incorporating broader global scholarship would strengthen the theoretical grounding of the study.

Is the study design appropriate and is the work technically sound?

Provide a clearer justification for analyzing only the 30 shortlisted candidates and discuss the implications of this sampling decision.

Although some methodological details are provided, additional clarification of analytical parameters and preprocessing decisions would improve reproducibility.

Are the conclusions drawn adequately supported by the results?

The conclusions generally align with the exploratory nature of the study; however, some claims regarding the usefulness of clustering for improving recruitment decision-making are somewhat stronger than the evidence supports. Because the analysis is based on a small, pre-screened sample and lacks external validation against actual hiring outcomes or job performance, the conclusions should emphasize the illustrative and exploratory nature of the findings rather than suggesting demonstrated improvements in recruitment effectiveness.

Is the work clearly and accurately presented and does it cite the current literature?

Partly

If applicable, is the statistical analysis and its interpretation appropriate?

I cannot comment. A qualified statistician is required.

Are all the source data underlying the results available to ensure full reproducibility?

No source data required

Is the study design appropriate and is the work technically sound?

Partly

Are the conclusions drawn adequately supported by the results?

Partly

Are sufficient details of methods and analysis provided to allow replication by others?

Yes

Reviewer Expertise:

Data Science

We confirm that we have read this submission and believe that we have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however we have significant reservations, as outlined above.

Ramdhani Wahyu Muhammad Educational Research and Evalu, Universitas Negeri Yogyakarta, Yogyakarta, Special Region of Yogyakarta, Indonesia

Competing interests: We declare that we have no competing interest on this work.

 7 5 2026

Dear Reviewers,

Thank you for your constructive evaluation of our manuscript and for approving the work with reservations. We have revised the manuscript substantially to address your comments.

1. Engagement with broader international literature

Comment:

The literature review focused primarily on clustering algorithms and included several regional or context-specific references. The manuscript would benefit from stronger engagement with more recent international literature on HR analytics, algorithmic decision-making in recruitment, and AI-assisted hiring systems.

Response:

We agree with this comment. In the revised manuscript, we have expanded the literature review by adding a new discussion of HR analytics, algorithm-assisted recruitment, AI-assisted hiring, fairness, transparency, explainability, adverse impact, and human oversight. We incorporated broader international literature and governance-oriented sources to strengthen the theoretical grounding of the study. The revised manuscript now positions K-Means clustering within the wider context of responsible recruitment analytics and decision support rather than only within clustering methodology.

Changes made:

We revised the Introduction and added a new literature review subsection on HR analytics and algorithm-assisted recruitment. We also added recent international references on AI-assisted recruitment, HR analytics, algorithmic fairness, and AI governance.

2. Justification for analysing only 30 shortlisted candidates

Comment:

Provide a clearer justification for analyzing only the 30 shortlisted candidates and discuss the implications of this sampling decision.

Response:

We agree. The revised manuscript now explains that 161 applicants initially applied, but only 30 candidates passed the company’s document-screening procedure and completed in-person competency testing. Only these 30 candidates had complete scores for all three assessment variables. We have clarified that this is a pre-screened analytical sample and that the findings should not be generalized to the full applicant pool or to construction applicants more broadly.

Changes made:

The Methods section now includes a clearer explanation of the screening procedure, the reason for limiting the analysis to 30 candidates, and the implications of this sampling decision for interpretation and generalizability.

3. Clarification of analytical parameters and preprocessing decisions

Comment:

Additional clarification of analytical parameters and preprocessing decisions would improve reproducibility.

Response:

We have revised the Methods section to clarify the preprocessing and clustering workflow. The revised manuscript states that all variables were measured on the same 0–100 scale, that raw scores were retained, that no additional normalization was applied, and that Euclidean squared distance was used in the clustering procedure. We also clarified that detailed clustering outputs and supporting materials are provided as extended data.

Changes made:

We revised the Methods section on assessment variables, preprocessing, outlier/sensitivity checks, and clustering procedure. We also updated the supplementary materials to align with the revised manuscript terminology.

4. Claims regarding usefulness of clustering for recruitment decision-making

Comment:

Some claims regarding the usefulness of clustering for improving recruitment decision-making were stronger than the evidence supports. Because the analysis is based on a small, pre-screened sample and lacks external validation against actual hiring outcomes or job performance, the conclusions should emphasize the illustrative and exploratory nature of the findings.

Response:

We agree and have revised the manuscript accordingly. The revised version no longer presents clustering as evidence of demonstrated improvement in recruitment effectiveness. Instead, clustering is framed as an exploratory and illustrative decision-support technique for organizing applicant assessment data. We explicitly acknowledge that the analysis was based on a small, pre-screened sample and was not externally validated against hiring outcomes or job performance.

Changes made:

We revised the Abstract, Results, Discussion, Limitations, and Conclusions to emphasize the exploratory nature of the findings. We also added explicit limitations regarding sample size, pre-screening, lack of external validation, and the need for future studies using larger datasets and job-performance outcomes.

We appreciate the reviewers’ comments, which helped us strengthen the theoretical framing, methodological transparency, and cautious interpretation of the manuscript.

 10.5256/f1000research.190104.r446627 Reviewer response for version 1 Gupta Deepak 1 Referee https://orcid.org/0000-0003-0069-7488 Penn State University, University Park, PA, USA

Competing interests: No competing interests were disclosed.

 3 2 2026 Copyright: © 2026 Gupta D 2026 This is an open access peer review report distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. recommendation approve-with-reservations

The manuscript addresses a relevant and practically important topic of using K‑Means clustering to support recruitment decisions in a construction consulting firm. The use of real organizational data, clear cluster descriptions, and intuitive 2D/3D visualizations makes the work accessible and potentially useful for practitioners interested in HR analytics and R-based implementations. However, the current version has substantial methodological and presentation limitations that need to be addressed before the article can be considered for indexing.

The study starts from 161 applicants but analyzes only 30 shortlisted candidates who “met the minimum requirements,” with no clear justification of the selection criteria or their implications. This design means the clustering is applied to a highly pre-filtered subset rather than the full applicant pool, which severely limits generalizability to “recruitment strategies” or “workforce selection” more broadly. The authors should explicitly describe the screening rules used to select the 30 candidates (cut‑offs, qualitative judgments, etc.). The authors should re-run the analysis on the full set of 161 applicants (if possible and ethically permissible) or reframe the paper as a methodological demonstration on a small, filtered sample and tone down claims about recruitment optimization.

Initial centroids are chosen manually as “Rejected = (60,75,85); Considered = (62,77,88); Accepted = (70,84,92)” with no methodological justification beyond being “chosen randomly,” which they are not. These values embed prior expectations about the three groups and risk steering the solution toward a desired structure, rather than letting the algorithm discover patterns in the data. The authors should justify the chosen centroids and make that explicit. In addition, run at least one standard initialization strategy (e.g., built‑in k-means in R with multiple random starts, or k-means++). The authors should also report whether the manually initialized solution is stable across different initializations; and if not, the conclusions about three clusters and their interpretation need to be more cautious.

The methods mention Euclidean distance and cite general clustering references, but there is no clear statement on whether the three variables were standardized or used in raw form. There is also no discussion of data cleaning beyond a vague reference to removing “irrelevant or incomplete data.” How many records, if any, were excluded and why? Whether any outliers were identified and how they were handled. The authors should explicitly state whether they standardized the variables before clustering; if not, justify why not and show at least descriptive ranges/variances. They should also provide a proper description of data-cleaning procedure. If outliers exist (some points in the 3D plot appear distant from cluster centers), please clarify whether they influence centroid placement and consider sensitivity analyses.

The discussion and conclusions repeatedly assert that K‑Means is “effective” and “improves recruitment decision-making,” yet there is no external validation. There is no comparison with actual hiring decisions made by the firm. No HR-expert assessment of whether clusters align with real performance or selection outcomes. If such data are not available, please qualify all claims about decision-making benefits as exploratory.

The author states in the “Underlying data” section that the dataset is not publicly available due to confidentiality and must be requested via the corresponding author. At the same time, the text refers to R scripts and supplementary materials in a Zenodo repository as extended data. The authors should clarify precisely what is in the Zenodo archive (R code, example data, figures, documentation) and referencing it consistently in the methods and data availability sections. If contractual and legal constraints allow, please consider publishing an anonymized dataset (at least the 30 shortlisted candidates; ideally all 161) with no identifying information. If this is impossible, explicitly justify the restriction and explain how others can still verify the methodology (e.g., via synthetic data included in the Zenodo package).

The conclusions and parts of the discussion are phrased quite strongly (e.g., “confirm that the K‑Means Clustering algorithm is an effective tool” and “this approach not only addresses challenges in evaluating applicants but also builds a foundation for sustainable talent management”), which is not fully justified given the small, preselected sample and lack of external validation.

Minor comments: The authors are suggested to modify the text in the abstract, discussion, and conclusions to reflect that findings are preliminary and illustrative for one firm and one small dataset.

The literature review is generally coherent but heavily focused on clustering and Indonesian/sector-specific references. There is relatively little engagement with the broader international work on AI in recruitment, algorithmic decision-making, or HR analytics.

Tables 2-7 list centroid updates and distances in excessive detail. It is pedagogically useful but makes the manuscript long and difficult to navigate. Some of the tables can be moved to the supplementary information with their summary in the main text.

There are some grammatical errors, please perform a proofread of the manuscript carefully.

Is the work clearly and accurately presented and does it cite the current literature?

Partly

If applicable, is the statistical analysis and its interpretation appropriate?

Partly

Are all the source data underlying the results available to ensure full reproducibility?

Partly

Is the study design appropriate and is the work technically sound?

Partly

Are the conclusions drawn adequately supported by the results?

Partly

Are sufficient details of methods and analysis provided to allow replication by others?

Partly

Reviewer Expertise:

Healthcare, 3D Printing, Artificial Intelligence

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.

Ramdhani Wahyu Muhammad Educational Research and Evalu, Universitas Negeri Yogyakarta, Yogyakarta, Special Region of Yogyakarta, Indonesia

Competing interests: The authors declare that they have no competing interests.

 6 2 2026

Response to Reviewer 2

Deepak Gupta (Penn State University, USA)

We thank the reviewer for the detailed and technically informed evaluation of our manuscript. The feedback has been instrumental in improving methodological rigor, transparency, and interpretive clarity. Our responses are provided below.

Comment 1: Pre-filtered sample and limited generalizability

Response:

We fully agree. The manuscript has been explicitly reframed to clarify that the clustering analysis was applied to a pre-screened subset of applicants who passed document screening. This design choice is now clearly justified in the Methods section, and all claims have been moderated accordingly. The study is presented as an exploratory analytical illustration rather than a full recruitment optimization model.

Comment 2: Manual centroid bias and need for robustness checks

Response:

This concern has been directly addressed. In addition to clarifying the rationale for initial centroid selection, we now report results from multiple random initializations (nstart = 50) using R’s kmeans() function. The three-cluster structure remained largely stable across runs, with only minor variation among borderline profiles. These findings are now reported and discussed in the revised manuscript.

Comment 3: Data normalization, outliers, and cleaning

Response:

The Methods section now explicitly states that all variables were measured on the same 0–100 scale, and therefore raw scores were used without normalization. We also added an outlier and sensitivity analysis, including a leave-one-out test removing the most distant observation. This test did not materially change the cluster structure, and validation metrics showed only marginal changes. These results are now reported to support robustness.

Comment 4: Lack of validation against real hiring outcomes

Response:

We explicitly acknowledge this limitation. The revised manuscript clearly states that no comparison with actual hiring decisions or post-employment performance was possible. All statements regarding recruitment support are now framed as illustrative and exploratory, not empirically validated improvements.

Comment 5: Data availability and Zenodo clarification

Response:

This issue has been resolved by clearly distinguishing:

Confidential underlying data (not publicly shareable);

Publicly available extended data hosted on Zenodo, including R scripts, figures, clustering outputs, and documentation.

 The Data availability section has been rewritten for clarity and compliance with journal policy.

Comment 6: Excessive tables and manuscript length

Response:

In line with the suggestion, detailed centroid iteration tables have been moved to extended data. The main text now focuses on final clustering outcomes, visual interpretation, and discussion, substantially improving readability.

Comment 7: Language and proofreading

Response:

The manuscript has been carefully proofread and revised to improve clarity, reduce repetition, and ensure consistent terminology throughout.

Final note:

We thank the reviewer for the rigorous and constructive critique. The revisions have significantly strengthened the manuscript’s technical soundness, transparency, and responsible positioning within data-driven recruitment research.

 10.5256/f1000research.190104.r446632 Reviewer response for version 1 Shaikh Sonia Najam 1 Referee https://orcid.org/0000-0001-7365-1064 Jiangsu University, Zhenjiang, China

Competing interests: No competing interests were disclosed.

 8 1 2026 Copyright: © 2026 Shaikh SN 2026 This is an open access peer review report distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. recommendation approve-with-reservations

The paper presents an interesting and practically valuable idea by using K-Means clustering to support recruitment decision-making in a construction firm. The study groups applicants into rejected, under consideration, and accepted categories based on AutoCAD skills, planning/supervision report ability, and adaptability, and it makes good use of R and visualization techniques to demonstrate the results. The topic is timely and relevant to HR analytics and industry needs, and overall the manuscript is readable, logically structured, and supported by real organizational data, which gives it strong applied significance.

However, while the work is clearly presented, it is only partly sufficient in terms of scientific rigor and presentation quality. The manuscript sometimes becomes repetitive and uses very long tables that make it difficult to follow the narrative smoothly. The literature review is decent, but it leans heavily on regional or context-specific studies, and it would benefit greatly from integrating more current international literature on AI-driven recruitment, algorithmic hiring, fairness, and HR analytics trends. Adding such perspectives would make the study more globally relevant and intellectually grounded.

In terms of study design, the concept is appropriate, but there are some important weaknesses that affect technical soundness. The biggest issue is that the study only analyzed 30 shortlisted candidates out of 161 applicants, without providing a clear justification. This introduces bias and limits how generalizable the findings are. Another concern is the selection of initial centroids, which appears subjective rather than determined using a standard procedure such as random initialization or k-means++. There is also no explanation of whether data normalization was applied, which is important since K-Means relies on Euclidean distance and differences in scale can significantly affect results. These issues really need to be addressed for the paper to be scientifically strong.

The paper partly allows replication because the clustering process is explained carefully and tables clearly trace centroid iterations and cluster formation. However, critical methodological details are missing. There is no detailed explanation of data cleaning, whether any applicants were excluded and why, how outliers were treated, whether ethical permission or company approval was obtained, and what settings were used in R. To make the study convincingly reproducible, the authors should provide anonymized raw data, the R script, and a clearer explanation of all methodological choices.

The analysis and interpretation are also only partly sufficient. Although clustering has been correctly applied and visualizations are very good, the study does not present objective validation of the clusters. There is no numerical reporting of validation metrics such as silhouette score, Davies–Bouldin index, or any form of comparison with real hiring outcomes or expert HR evaluation. Without such validation, the findings are mostly descriptive, and the claim that clustering improves recruitment decision-making is not empirically proven. Adding cluster validity results and some form of comparison or evaluation would significantly strengthen the credibility of the conclusions.

Data availability is partly fulfilled. While applicant scores are shown in tables, there is no formal dataset sharing, no supplementary files, and no data availability statement. Reproducibility is an important expectation, so providing anonymized data and the analysis code will help readers trust and build upon the work.

The conclusions are relevant and aligned with the results, but they are currently expressed too strongly considering the methodological limitations and lack of validation. The study needs to acknowledge its limitations, including the small selective sample, potential bias, lack of external validation, and ethical concerns such as fairness and transparency in automated hiring. The tone of the conclusions should be slightly moderated to reflect that the findings are promising but not definitively proven.

Overall, this is a promising and meaningful study with clear practical relevance, strong visualization, and a good foundation. However, to make it scientifically sound and suitable for indexing, the authors must justify or expand the sample, apply standard centroid initialization or explain theirs clearly, normalize or justify scale handling, provide quantitative cluster validation, share data and code, strengthen literature support, and include a more explicit discussion of limitations and ethical considerations. 

Is the work clearly and accurately presented and does it cite the current literature?

Partly

If applicable, is the statistical analysis and its interpretation appropriate?

Partly

Are all the source data underlying the results available to ensure full reproducibility?

Partly

Is the study design appropriate and is the work technically sound?

Partly

Are the conclusions drawn adequately supported by the results?

Partly

Are sufficient details of methods and analysis provided to allow replication by others?

Partly

Reviewer Expertise:

Human Resource Management, Artificial intelligence, Big data analytics capability

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.

Ramdhani Wahyu Muhammad Educational Research and Evalu, Universitas Negeri Yogyakarta, Yogyakarta, Special Region of Yogyakarta, Indonesia

Competing interests: The authors declare that they have no competing interests.

 6 2 2026

Response to Reviewer 1

Sonia Najam Shaikh (Jiangsu University, China)

We sincerely thank the reviewer for the careful reading of our manuscript and for the constructive and detailed feedback. We have revised the manuscript substantially to address all major concerns raised. Our point-by-point responses are provided below.

Comment 1: Limited international literature and lack of broader AI/HR analytics perspective

Response:

Thank you for this important suggestion. We have expanded the literature review and discussion to better situate the study within international debates on data-driven recruitment, employability, and HR analytics. Specifically, we incorporated recent global literature addressing analytical decision support, adaptability, and career sustainability (e.g., Akkermans et al., 2024; Donald et al., 2024; Rawat et al., 2024; Zhang et al., 2024; Van der Heijden et al., 2024).

In addition, we explicitly positioned clustering as an exploratory decision-support tool, aligned with international guidance on responsible and transparent use of analytics in employment contexts.

Comment 2: Analysis limited to 30 candidates out of 161 applicants

Response:

We fully agree that this required clearer justification. The revised manuscript now explicitly explains that the clustering analysis was conducted on 30 candidates who passed formal document screening conducted by the company as part of standard recruitment procedures. We clearly state that the study analyzes a pre-screened assessment sample, not the full applicant pool.

Accordingly, the study has been reframed as an exploratory methodological illustration rather than a comprehensive recruitment optimization model. Limitations related to selection bias and generalizability are now explicitly acknowledged in the Methods, Results, and Conclusions sections.

Comment 3: Subjective centroid selection and lack of standard initialization

Response:

Thank you for highlighting this critical methodological issue. We have clarified that the manually specified initial centroids were derived from preliminary inspection of score distributions and used for transparency and pedagogical traceability.

To address concerns regarding subjectivity and robustness, we have now:

Added a dedicated subsection describing multiple random initializations using the built-in kmeans() function in R (nstart = 50);

Reported stability checks, showing that the three-cluster structure was largely consistent across runs, with only minor variations among borderline cases;

Moderated all interpretations to emphasize that clusters reflect similarity patterns, not predefined hiring categories.

Comment 4: Missing details on data cleaning, ethics, and reproducibility

Response:

This concern has been fully addressed. The revised manuscript now includes:

A clearer description of document screening and data preprocessing;

Explicit confirmation that all analyzed records were complete and measured on the same 0–100 scale;

A detailed Ethical approval and informed consent section clarifying the use of fully anonymized secondary data under a formal Data Usage Agreement;

An updated Data availability statement;

An updated Zenodo repository containing R scripts, clustering outputs, and documentation to support reproducibility.

Comment 5: Lack of objective cluster validation

Response:

We agree with this observation. To strengthen technical rigor, we have added a dedicated cluster validity assessment section reporting internal validation metrics. Specifically, we report:

Mean silhouette coefficient (0.16);

Davies–Bouldin Index (DBI = 2.05).

 We explicitly interpret these values as diagnostic indicators of modest separation, appropriate for exploratory analysis, and avoid presenting them as evidence of predictive effectiveness.

Comment 6: Overly strong conclusions

Response:

The Conclusions section has been revised to adopt a more cautious and balanced tone. Claims regarding effectiveness and decision-making improvement are now framed as exploratory and context-specific. Limitations related to sample size, pre-screening, lack of external validation, and ethical considerations are explicitly acknowledged.

Final note:

We sincerely appreciate the reviewer’s thoughtful comments, which have significantly strengthened the manuscript’s methodological transparency, analytical caution, and international relevance.