Analysis and Prediction of At-Risk Students Using Machine Learning Algorithms

Hamid Salarian; Soheila Gheisari

arxiv: 2606.20617 · v1 · pith:7I4ATG6B · submitted 2026-05-24 · cs.CY · cs.LG

Analysis and Prediction of At-Risk Students Using Machine Learning Algorithms

Soheila Gheisari , Hamid Salarian This is my paper

Reviewed by Pith2026-06-29 23:49 UTCgrok-4.3pith:7I4ATG6B open to challenge →

classification cs.CY cs.LG

keywords machine learningstudent attritionat-risk predictionlogistic regressionsupport vector machineshigher education retentionpredictive modelingeducational data mining

0 comments

The pith

Logistic regression and linear SVM achieve the highest accuracy when predicting which students will withdraw using academic performance, demographic data, and enrollment records.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper sets out to test whether machine learning can identify students likely to leave their programs before they do so. It trains and compares four standard classifiers on a mix of grades, background variables, and enrollment history, then reports which ones flag at-risk cases most reliably. A reader would care because early, accurate flags let schools offer targeted support that might keep students enrolled and protect institutional revenue. The work therefore positions routine student data as a practical input for retention decisions rather than a post-hoc record.

Core claim

When the four models were evaluated on the student dataset, logistic regression and linear SVM returned the highest accuracy, showing that machine learning techniques can be applied to academic performance, demographic data, and enrollment records to detect students at risk of withdrawal.

What carries the argument

Supervised classification models (logistic regression, random forest, SVM, KNN) trained to output a binary risk label from student records.

If this is right

Schools could run these models each term to produce a list of students for early outreach.
The same data pipeline could be reused to monitor changes in risk scores after support interventions are applied.
Resource planning could shift from uniform services to targeted allocation based on the model outputs.
Enrollment offices could incorporate risk scores into admission or advising workflows.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The approach may need periodic retraining if student demographics or program structures change over time.
Accuracy alone does not guarantee that flagged students will respond to the interventions that schools actually offer.
Extending the feature set to include engagement metrics such as login frequency could improve detection without new data sources.

Load-bearing premise

The collected student records are representative of the broader population of interest and free of systematic biases that would make the learned patterns fail on new students.

What would settle it

Retraining the same models on a later or different cohort of students and finding that accuracy drops substantially below the levels reported in the study.

read the original abstract

Student attrition represents a significant challenge for higher education institutions because it impacts both academic results and financial viability. Machine learning provides an effective solution to identify students who require assistance before they leave their academic programs. The research investigates how machine learning approaches enable institutions to predict student withdrawal and enrollment cancellation through data-driven insights for strategic decisionmaking. The evaluation of models includes Logistic Regression, Random Forest, Support Vector Machines (SVM), and K-Nearest Neighbors (KNN) based on academic performance and demographic data and enrollment records. The results show that logistic regression and linear SVM models produced the highest accuracy which demonstrates ML's capability to detect students at risk.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Routine application of standard classifiers to student attrition with no experimental details, so the accuracy claim cannot be checked.

read the letter

The main takeaway is that this paper cannot be properly evaluated because the abstract provides no experimental details whatsoever. It claims logistic regression and linear SVM achieved the highest accuracy on academic, demographic, and enrollment data for predicting at-risk students, but without numbers, sample sizes, validation methods, or baselines, that claim is impossible to check.

The work applies four common classifiers—Logistic Regression, Random Forest, SVM, and KNN—to a known problem in higher education. Student attrition is indeed a real issue for institutions, affecting both academic outcomes and finances, and using ML for early detection makes sense in principle. The paper does a decent job of framing that practical need.

However, there is nothing new in the approach. These models are standard, the data types are typical for educational data mining, and no new algorithms or insights are presented. The abstract cites no prior work or comparisons, which makes it hard to see what this adds.

The soft spot is the total lack of methodological transparency. No mention of how the data was prepared, split for training and testing, or any performance metrics beyond the vague "highest accuracy." This is not a small omission; it prevents any assessment of whether the results are reliable or generalizable. The reader's weakest assumption about the dataset being representative cannot be checked either.

This paper would mainly interest practitioners in university administration looking for basic ideas on applying ML to retention, but it offers little for researchers or anyone needing solid evidence. It does not reorganize thinking in the field or provide reproducible findings.

I would not bring this to a reading group, and I would not cite it. It does not deserve peer review in this state; the authors would need to include full results, comparisons, and details before it could be considered seriously.

Referee Report

1 major / 2 minor

Summary. The manuscript analyzes the application of machine learning algorithms—Logistic Regression, Random Forest, Support Vector Machines (SVM), and K-Nearest Neighbors (KNN)—to predict at-risk students based on academic performance, demographic data, and enrollment records. The key finding is that logistic regression and linear SVM models achieved the highest accuracy, illustrating the potential of ML techniques for early detection of students likely to withdraw or cancel enrollment.

Significance. Should the empirical results prove robust upon detailed examination, this work could contribute meaningfully to the field of educational data mining by offering practical insights for higher education institutions to implement proactive support systems, thereby addressing student attrition challenges that affect academic and financial outcomes.

major comments (1)

Abstract: The abstract states that logistic regression and linear SVM produced the highest accuracy but supplies no sample size, validation method, baseline comparison, quantitative accuracy scores, or error bars. This absence prevents evaluation of the central claim that these models outperform the others and demonstrate ML's capability in this domain.

minor comments (2)

Abstract: The term 'decisionmaking' is missing a hyphen and should read 'decision-making'.
Abstract: The abstract sentence describing the evaluation is lengthy and could be split for improved readability.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their review and constructive feedback. We address the single major comment below and will revise the manuscript accordingly.

read point-by-point responses

Referee: Abstract: The abstract states that logistic regression and linear SVM produced the highest accuracy but supplies no sample size, validation method, baseline comparison, quantitative accuracy scores, or error bars. This absence prevents evaluation of the central claim that these models outperform the others and demonstrate ML's capability in this domain.

Authors: We agree that the abstract would be strengthened by including these quantitative details. The full manuscript reports a dataset of student records, 5-fold cross-validation, accuracy scores for all four classifiers (with logistic regression and linear SVM highest), and standard deviations across folds. In the revised version we will condense these elements into the abstract while preserving its length constraints, enabling readers to evaluate the central claim directly. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper applies standard supervised classification algorithms (Logistic Regression, Random Forest, SVM, KNN) to a dataset of academic, demographic, and enrollment features to predict student attrition. No equations, derivations, fitted parameters renamed as predictions, or self-citations appear in the provided abstract or described content. The reported accuracies are direct empirical outcomes of model training and evaluation on the input data, with no reduction of any claimed result to its own inputs by construction. This is a conventional applied ML study whose central claim rests on external data and standard algorithms rather than any self-referential loop.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The abstract invokes the standard supervised-learning assumption that the collected student records are i.i.d. samples from the population of interest; no free parameters, invented entities, or additional axioms are stated.

axioms (1)

domain assumption Student records used for training are independent and identically distributed samples from the target population.
Implicit in any claim that a fitted classifier will generalize to future students.

pith-pipeline@v0.9.1-grok · 5628 in / 1097 out tokens · 20897 ms · 2026-06-29T23:49:57.344804+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

9 extracted references · 2 canonical work pages

[1]

Internal corporate responsibility as a legitimacy strategy for branding and employee retention: A perspective of higher education institutions,

A. Ikram, M. Fiaz, A. Mahmood, A. Ahmad, and R. Ashfaq, “Internal corporate responsibility as a legitimacy strategy for branding and employee retention: A perspective of higher education institutions,” Journal of Open Innovation: Technology, Market, and Complexity, vol. 7, no. 1, 2021

2021
[2]

Improving student retention in institutions of higher education through machine learning: A sustainable approach,

W. Villegas -Ch, J. Govea, and S. Revelo -Tapia, “Improving student retention in institutions of higher education through machine learning: A sustainable approach,” Sustainability, vol. 15, no. 19, 2023

2023
[3]

Total rewards and retention: Case study of higher education institutions in pakistan,

C. S. Akhtar, A. Aamir, M. A. Khurshid, M. M. Q. Abro, and J. Hussain, “Total rewards and retention: Case study of higher education institutions in pakistan,” Procedia - Social and Behavioral Sciences, vol.210, pp. 251 –259, 2015, proceedings of the 4th International Conference on Leadership, Technology, Innovation and Business Management (ICLTIBM-2014)

2015
[4]

Machine learning model and ensemble algorithm for prediction of students’ retention and graduation in education,

K. Okoye, J. T. Nganji, J. Escamilla, and S. Hosseini, “Machine learning model and ensemble algorithm for prediction of students’ retention and graduation in education,” Computers and Education:Artificial Intelligence, vol. 6, p. 100205, 2024

2024
[5]

A systematic review on the deployment and effectiveness of data analytics in higher education to improve student outcomes,

C. Foster and P. Francis, “A systematic review on the deployment and effectiveness of data analytics in higher education to improve student outcomes,” Assessment & Evaluation in Higher Education, vol. 45, no. 6, pp. 822–841, 2020

2020
[6]

Employee Retention and Job Performance Attributes in Private Institutions of Higher Education,

C. C. Y. L. S. Teng, “Employee Retention and Job Performance Attributes in Private Institutions of Higher Education,” International Journal of Business and Administrative Studies, vol. 3, no. 5, pp. 158–165, 10 2017

2017
[7]

Comparative Study of Machine Learning Algorithms for Student Retention, early warning and intervention systems for Institutions of Higher Learning,

M. C. I. Madahana and J. E. D. Ekoru, "Comparative Study of Machine Learning Algorithms for Student Retention, early warning and intervention systems for Institutions of Higher Learning," 2024 IEEE 15th Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON), Yorktown Heights, NY, USA, 2024, pp. 366-371, doi: 10.1109/UEMCON6287...

work page doi:10.1109/uemcon62879.2024.10754680 2024
[8]

Forecasting Student Attrition Using Machine Learning

Prajwal, P., L. R. Sahana, and V. Kanchana. "Forecasting Student Attrition Using Machine Learning." 2024 4th Asian Conference on Innovation in Technology (ASIANCON). IEEE, 2024

2024
[9]

Student Retention Model via Machine Learning and Predictive analysis,

M. M. Malatji, R. Mohlomi, G. Kirui, S. Mndebele, J. E. D. Ekoru and M. C. I. Madahana, "Student Retention Model via Machine Learning and Predictive analysis," 2024 IEEE World AI IoT Congress (AIIoT), Seattle, WA, USA, 2024, pp. 212 -218, doi: 10.1109/AIIoT61789.2024.10578977

work page doi:10.1109/aiiot61789.2024.10578977 2024

[1] [1]

Internal corporate responsibility as a legitimacy strategy for branding and employee retention: A perspective of higher education institutions,

A. Ikram, M. Fiaz, A. Mahmood, A. Ahmad, and R. Ashfaq, “Internal corporate responsibility as a legitimacy strategy for branding and employee retention: A perspective of higher education institutions,” Journal of Open Innovation: Technology, Market, and Complexity, vol. 7, no. 1, 2021

2021

[2] [2]

Improving student retention in institutions of higher education through machine learning: A sustainable approach,

W. Villegas -Ch, J. Govea, and S. Revelo -Tapia, “Improving student retention in institutions of higher education through machine learning: A sustainable approach,” Sustainability, vol. 15, no. 19, 2023

2023

[3] [3]

Total rewards and retention: Case study of higher education institutions in pakistan,

C. S. Akhtar, A. Aamir, M. A. Khurshid, M. M. Q. Abro, and J. Hussain, “Total rewards and retention: Case study of higher education institutions in pakistan,” Procedia - Social and Behavioral Sciences, vol.210, pp. 251 –259, 2015, proceedings of the 4th International Conference on Leadership, Technology, Innovation and Business Management (ICLTIBM-2014)

2015

[4] [4]

Machine learning model and ensemble algorithm for prediction of students’ retention and graduation in education,

K. Okoye, J. T. Nganji, J. Escamilla, and S. Hosseini, “Machine learning model and ensemble algorithm for prediction of students’ retention and graduation in education,” Computers and Education:Artificial Intelligence, vol. 6, p. 100205, 2024

2024

[5] [5]

A systematic review on the deployment and effectiveness of data analytics in higher education to improve student outcomes,

C. Foster and P. Francis, “A systematic review on the deployment and effectiveness of data analytics in higher education to improve student outcomes,” Assessment & Evaluation in Higher Education, vol. 45, no. 6, pp. 822–841, 2020

2020

[6] [6]

Employee Retention and Job Performance Attributes in Private Institutions of Higher Education,

C. C. Y. L. S. Teng, “Employee Retention and Job Performance Attributes in Private Institutions of Higher Education,” International Journal of Business and Administrative Studies, vol. 3, no. 5, pp. 158–165, 10 2017

2017

[7] [7]

Comparative Study of Machine Learning Algorithms for Student Retention, early warning and intervention systems for Institutions of Higher Learning,

M. C. I. Madahana and J. E. D. Ekoru, "Comparative Study of Machine Learning Algorithms for Student Retention, early warning and intervention systems for Institutions of Higher Learning," 2024 IEEE 15th Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON), Yorktown Heights, NY, USA, 2024, pp. 366-371, doi: 10.1109/UEMCON6287...

work page doi:10.1109/uemcon62879.2024.10754680 2024

[8] [8]

Forecasting Student Attrition Using Machine Learning

Prajwal, P., L. R. Sahana, and V. Kanchana. "Forecasting Student Attrition Using Machine Learning." 2024 4th Asian Conference on Innovation in Technology (ASIANCON). IEEE, 2024

2024

[9] [9]

Student Retention Model via Machine Learning and Predictive analysis,

M. M. Malatji, R. Mohlomi, G. Kirui, S. Mndebele, J. E. D. Ekoru and M. C. I. Madahana, "Student Retention Model via Machine Learning and Predictive analysis," 2024 IEEE World AI IoT Congress (AIIoT), Seattle, WA, USA, 2024, pp. 212 -218, doi: 10.1109/AIIoT61789.2024.10578977

work page doi:10.1109/aiiot61789.2024.10578977 2024