Pith. sign in

REVIEW 4 major objections 5 minor 37 references

Social Media Sentiments Analysis on the July Revolution in Bangladesh: A Hybrid Transformer Based Machine Learning Approach

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper claims that merging the sentence-level embeddings of BanglaBERT, mBERT, and XLM-RoBERTa into a hybrid feature vector called XMB-BERT, then reducing with PCA and classifying with a Voting Classifier, reaches 83.7% accuracy on…

desk verdict A genuinely useful new Bangla dataset wrapped in a thin hybrid-embedding benchmark whose headline number is probably inflated by a train/test leakage. read the letter →

arxiv 2507.11084 v1 pith:BCTRLT76 submitted 2025-07-15 cs.CL

Md. Sabbir Hossen , Md. Saiduzzaman , Pabon Shaha This is my paper
classification cs.CL
keywords sentimentanalysisBanglaNLPtransformermodelshybridfeatureextractionXMB-BERTJulyRevolutionsocialmediavotingclassifier
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to show that combining the sentence-level embeddings of three pretrained transformers—BanglaBERT, mBERT, and XLM-RoBERTa—into a single hybrid feature vector improves sentiment classification for Bangla social media text about the July Revolution. On a new manually labeled dataset of 4,200 comments, the hybrid feature set with a Voting Classifier reaches 83.7% accuracy, above any single transformer feature set. The result matters because Bangla is a low-resource language and political discourse on social media is a hard, understudied target. If the claim holds, hybrid feature concatenation plus PCA plus ensemble classification is a viable recipe for sentiment analysis in low-resource, politically charged settings.

What carries the argument

The central object is the hybrid feature extractor named XMB-BERT: the [CLS] token embeddings—one sentence-level summary vector per comment—from XLM-RoBERTa, mBERT, and BanglaBERT are concatenated into a single vector. PCA then compresses that high-dimensional vector, and a Voting Classifier, which combines the predictions of several base classifiers, serves as the classification head. The mechanism is complementary representation: the paper argues that the three transformers capture different linguistic signals, so the fused vector encodes more sentiment-relevant variation than any single model's embeddings.

What would settle it

Have two or more independent annotators label the same 4,200 comments and compute inter-annotator agreement, then train XMB-BERT on one annotator's labels and test on another's; if agreement is low or accuracy drops sharply, the 83.7% measures the authors' labeling criteria rather than a generalizable property of the text.

Watch

Extended reading notes

Core claim

The paper reports that a hybrid feature extractor it calls XMB-BERT, formed by concatenating the [CLS] sentence embeddings of mBERT, XLM-RoBERTa, and BanglaBERT, gives better sentiment classification than any of those three transformers alone. With principal component analysis for dimensionality reduction and a Voting Classifier as the final model, the system reaches 83.7% accuracy, 84.1% precision, 83.7% recall, and an 83.7% F1-score on a balanced dataset of 4,200 Bangla comments about the July Revolution. The authors also claim this is the first sentiment-analysis benchmark for Bangla-language social media text on the July Revolution and release the labeled dataset publicly for further research.

Load-bearing premise

The load-bearing premise is that the authors' manual labels of positive, negative, and neutral in Section 3.1 are accurate and consistent ground truth, even though the paper notes they are susceptible to annotator bias and no formal fact-checking was performed.

Editorial extensions

If this is right

  • With XMB-BERT features, the Voting Classifier reaches 83.7% accuracy, the best result among the feature-extractor and classifier combinations the paper tested.
  • The hybrid feature set beats each of BanglaBERT, mBERT, and XLM-RoBERTa when compared under the same classifiers in most pairings, so the gain is attributed to feature fusion rather than to any one transformer.
  • The publicly released 4,200-comment dataset gives other researchers a labeled benchmark for Bangla political sentiment, the first tied to the July Revolution.
  • PCA keeps the concatenated embeddings computationally manageable, so the hybrid approach runs on modest hardware, although training still takes more than three hours.
  • The ROC analysis shows the strongest class separability for neutral comments, with an AUC of 0.96, suggesting the model is most confident at distinguishing neutral from emotionally charged text.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A test the paper does not run: have several independent annotators relabel the same 4,200 comments and measure inter-annotator agreement; without that, the 83.7% accuracy could reflect the authors' own labeling criteria rather than a property that transfers to other raters.
  • The same concatenation recipe could be tried on other low-resource languages that have a monolingual BERT variant; the paper's logic predicts the union of multilingual and language-specific embeddings will beat either alone, but that extension is not tested here.
  • Because the dataset excludes code-mixed and noisy text, the reported margin over single models may shrink on raw, uncurated social media content; feeding unprocessed comments into XMB-BERT would reveal how much of the gain depends on preprocessing.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper presents a sentiment-analysis benchmark for Bangla-language social media comments about the July Revolution in Bangladesh. The authors introduce a new manually labeled, balanced dataset of 4,200 comments and propose Hybrid XMB-BERT, a feature extractor formed by concatenating the [CLS] embeddings of mBERT, XLM-RoBERTa, and BanglaBERT, followed by PCA dimensionality reduction and classification with eleven machine learning classifiers. The headline result is that XMB-BERT with a voting classifier achieves 83.7% accuracy, which the paper claims outperforms all other feature-extraction/classifier combinations. The paper also claims to be the first sentiment-analysis study of Bangla social media text specifically about the July Revolution.

Significance. If the evaluation were sound, the paper would make a useful empirical contribution: a publicly released, balanced dataset for an under-resourced language and a politically significant event, plus a systematic comparison of several transformer feature extractors and eleven classifiers. The dataset release (DOI provided) is a concrete asset, and the comparison across many configurations is informative even where the proposed hybrid is not uniformly best. However, the central quantitative claim is currently compromised by a preprocessing-leakage issue in the described pipeline and by unsupported 'outperforms all' statements that are contradicted by the paper's own table. The significance of the result therefore depends on a corrected evaluation and more careful statistical reporting.

major comments (4)
  1. [§4.2, Algorithm 1] The implementation procedure applies dimensionality reduction before the train/test split: the loop sets XDR ← DR[j](XV) on the full embedded set, and only then performs Train-Test Split (80:20)(XDR, Y). This means PCA is fit on all 4,200 samples, including the 840 test instances, before any held-out set exists. This contradicts Section 3.3, which states that the dataset is partitioned before feature extraction. In high-dimensional [CLS] embedding spaces, PCA axes fit on the pooled data can absorb variance from test samples and make the transformed test features artificially consistent with the training distribution, biasing the reported 83.7% accuracy upward. The protocol must be corrected (fit PCA only on the training fold and apply the learned transformation to the test fold, or use a cross-validation pipeline), and all results must be recomputed. The same loop also feeds t-SNE and UMAP embeddings into classifiers even though Section 3.5 describes them as visualization only; if they were used for classification, they are subject to the same leakage, and if not, the pseudocode should be revised to remove them from the classifier loop.
  2. [§5.1, Table II] The text claims that XMB-BERT 'outperforms the other models across all metrics and classifiers,' but Table II contradicts this. For example, with Logistic Regression, mBERT achieves 76.3% accuracy versus 75.6% for XMB-BERT; with SVM, XLM-RoBERTa achieves 78.2% versus 74.1% for XMB-BERT; and with KNN, BanglaBERT achieves 75.4% versus 73.4% for XMB-BERT. The claim should be restricted to specific classifiers (voting classifier, XGBoost, LGBM, Gradient Boosting) and the narrative revised to avoid a global superiority statement.
  3. [§5.1, Table II] The headline gain of XMB-BERT with the voting classifier (83.7%) over the best single-transformer configuration (XLM-RoBERTa with LGBM, 82.2%) is 1.5 percentage points on 840 test instances. No confidence intervals, significance tests, or repeated-seed variance are reported, and the identity of the best classifier changes across feature sets. The claimed superiority should be supported by a paired test (e.g., McNemar's test) or by reporting the variance across multiple random splits or training seeds.
  4. [§5.2, Limitations] The paper correctly acknowledges that the manual labels are 'susceptible to annotator bias' and that no formal fact-checking was performed. Because the entire evaluation relies on this ground truth, the absence of inter-annotator agreement (e.g., Cohen's kappa) or an independent validation set means the reported accuracy may partially measure the authors' labeling criteria rather than generalizable sentiment. Please add a small annotation study or at least an explicit statement of how label disagreements would affect the reported results.
minor comments (5)
  1. [§3.6 and §4.2, Algorithm 1] The text says 11 classifiers were used, but the list in Algorithm 1 contains only 10 entries and omits Gradient Boosting, which does appear in Table II. Please make the notation consistent.
  2. [Abstract and §3.5] 'Principle Component Analysis' should be 'Principal Component Analysis'.
  3. [§3.1] The collection period is stated as 'September and October' without a year; specify 2024 for clarity.
  4. [§3.5 and Algorithm 1] The role of t-SNE and UMAP is described as visualization only, but Algorithm 1 includes them in the DR loop that feeds classifiers; clarify whether classifier results were computed on t-SNE/UMAP embeddings or only on PCA, and revise the pseudocode accordingly.
  5. [References] Several self-citations (e.g., [24], [28], [30], [34], [37]) refer to unrelated tasks such as strawberry leaf disease, cardiovascular risk, and jellyfish identification; these do not support the sentiment-analysis methodology and could be removed or replaced with directly relevant prior work.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation found; evaluation leakage and label subjectivity are validity issues, not circularity.

full rationale

The paper's central claim is an empirical benchmark result: a hybrid XMB-BERT feature set plus voting classifier reaches 83.7% accuracy on a manually labeled dataset. There is no derivation chain in which a predicted quantity is defined in terms of itself, no fitted parameter is renamed as a prediction, and no load-bearing conclusion rests on a self-citation. The self-citations in Sections 3.6 and 4.1 (references [24], [28], [30], [34], [37]) are ordinary citations for standard algorithms such as SVM, KNN, XGBoost, and voting classifiers; they do not supply the central claim. The pre-trained transformers (mBERT, XLM-RoBERTa, BanglaBERT) are external artifacts, and the model is evaluated on 840 test instances held out from the 4,200-sample set. The most notable methodological concern is Algorithm 1 in Section 4.2, which applies PCA to the full embedded dataset before the 80:20 split, meaning the PCA transform is fit on data that includes the test samples; this is a data-leakage/evaluation-protocol flaw that can inflate reported accuracy, but it does not make the 83.7% accuracy equivalent to an input by construction or otherwise constitute circular reasoning under the definitions used here. Similarly, the authors' acknowledged limitation that labels are susceptible to annotator bias and were not fact-checked is a ground-truth validity concern, not a circularity. Because the benchmark is self-contained and the claimed result is not forced by its own inputs or by a self-citation chain, the circularity score is 0.

Assumptions & free parameters 3 free parameters · 5 assumptions · 1 invented entities

The central claim rests on label quality, representativeness of the manual sample, the informativeness of frozen transformer embeddings, the safety of the preprocessing, and the assumption that applying PCA before the split does not bias results. The empirical nature means there is no theoretical derivation, but several hidden configuration choices act as free parameters.

free parameters (3)
  • PCA component count or variance retained
    Not reported; this choice affects all downstream classifier results.
  • Voting classifier composition and weights
    The voting classifier's base estimators and weighting are unspecified, so the headline result depends on an undisclosed configuration.
  • Classifier hyperparameters
    No hyperparameter values are reported for the eleven classifiers; default or arbitrary settings may drive the comparison.
assumptions (5)
  • domain assumption Author-assigned sentiment labels are treated as ground truth.
    Section 3.1 defines labeling criteria; the limitations section admits annotator bias and the absence of formal fact-checking.
  • domain assumption The 4,200 manually collected comments are representative of public sentiment during and after the July Revolution.
    Section 3.1 describes manual collection from unspecified public posts without a documented sampling protocol.
  • domain assumption Pre-trained transformer [CLS] embeddings encode sentiment-relevant features for Bangla social media text.
    Section 3.4 uses frozen embeddings with no fine-tuning or task-specific validation.
  • domain assumption Preprocessing (stopword removal, stemming, normalization) preserves sentiment-bearing content.
    Section 3.2 applies aggressive cleaning; no ablation or error analysis is provided.
  • ad hoc to paper Applying PCA before the train/test split does not leak test information into training.
    Algorithm 1 applies dimensionality reduction before Train-Test Split, so the test set contributes to the PCA projection.
invented entities (1)
  • Hybrid XMB-BERT
    purpose: A feature vector made by concatenating the [CLS] embeddings of mBERT, XLM-RoBERTa, and BanglaBERT.
    It is a named pipeline of existing components; the only evidence is the benchmark reported in this paper, with no external validation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Social Media Sentiments Analysis on the July Revolution in Bangladesh: A Hybrid Transformer Based Machine Learning Approach." pith.science (2026). https://pith.science/paper/BCTRLT76

@misc{pith2026250711084,
  author       = {Pith},
  title        = {Pith review of: Social Media Sentiments Analysis on the July Revolution in Bangladesh: A Hybrid Transformer Based Machine Learning Approach},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BCTRLT76}},
  note         = {Machine review of arXiv:2507.11084}
}
read the original abstract

The July Revolution in Bangladesh marked a significant student-led mass uprising, uniting people across the nation to demand justice, accountability, and systemic reform. Social media platforms played a pivotal role in amplifying public sentiment and shaping discourse during this historic mass uprising. In this study, we present a hybrid transformer-based sentiment analysis framework to decode public opinion expressed in social media comments during and after the revolution. We used a brand new dataset of 4,200 Bangla comments collected from social media. The framework employs advanced transformer-based feature extraction techniques, including BanglaBERT, mBERT, XLM-RoBERTa, and the proposed hybrid XMB-BERT, to capture nuanced patterns in textual data. Principle Component Analysis (PCA) were utilized for dimensionality reduction to enhance computational efficiency. We explored eleven traditional and advanced machine learning classifiers for identifying sentiments. The proposed hybrid XMB-BERT with the voting classifier achieved an exceptional accuracy of 83.7% and outperform other model classifier combinations. This study underscores the potential of machine learning techniques to analyze social sentiment in low-resource languages like Bangla.

Figures

Figures reproduced from arXiv: 2507.11084 by the authors.

Figure 1
Figure 1. Graffiti of Revolution and Unity: Visual Narratives of the July [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Proposed Workflow Diagram for Sentiment Analysis [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 5
Figure 5. Distribution of text lengths in the dataset, indicating a right-skewed [PITH_FULL_IMAGE:figures/full_fig_p003_5.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Sample comments from the collected dataset, along with their English [PITH_FULL_IMAGE:figures/full_fig_p003_4.png]
Figure 6
Figure 6. Figure 6: t-distributed Stochastic Neighbor Embedding (t-SNE) visualize the data before dimensionality reduction. (a) t-SNE visualization of BanglaBERT, [PITH_FULL_IMAGE:figures/full_fig_p004_6.png]
Figure 7
Figure 7. Figure 7: Uniform Manifold Approximation and Projection (UMAP) visualize data before dimensionality reduction. (a) UMAP visualization of BanglaBERT, [PITH_FULL_IMAGE:figures/full_fig_p004_7.png]
Figure 8
Figure 8. Figure 8: Principle Component Analysis (PCA) visualizes the high dimensional data after dimensionality reduction. (a) PCA visualization of BanglaBERT, [PITH_FULL_IMAGE:figures/full_fig_p004_8.png]
Figure 9
Figure 9. Figure 9: The Architecture of the proposed Hybrid XMB-BERT. It shows that preprocessed data is fed into three transformer models: mBERT, BanglaBERT, [PITH_FULL_IMAGE:figures/full_fig_p005_9.png]
Figure 10
Figure 10. Figure 10: Confusion matrices compare the performance of three different feature extraction techniques with Proposed Hybrid XMB-BERT: (a) BanglaBERT [PITH_FULL_IMAGE:figures/full_fig_p006_10.png]
Figure 11
Figure 11. Figure 11: Receiver-operating characteristic curve compare the performance of three different feature extraction techniques with Proposed Hybrid XMB-BERT: [PITH_FULL_IMAGE:figures/full_fig_p006_11.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

37 extracted references · 29 canonical work pages

  1. [1]

    Prosecuting the perpetrators of july- august massacre: A case for restorative justice,

    M. Asadullah and N. S. Ritu, “Prosecuting the perpetrators of july- august massacre: A case for restorative justice,” The Daily Star , 2024. Availaable: https://bit.ly/40sInhD

    work page 2024

  2. [2]

    The role of political parties in bangladesh’s july revolution of 2024: Insights from sufi perspectives,

    S. R. H. Chowdury, “The role of political parties in bangladesh’s july revolution of 2024: Insights from sufi perspectives,” International Journal of Research and Innovation in Social Science , vol. 8, no. 11, pp. 2077–2093, 2024

    work page 2024

  3. [3]

    July massacre of 2024,

    B. Data, “July massacre of 2024,” Bangladesh Data, 2024. Available: https://bit.ly/42a9QGh

    work page 2024

  4. [4]

    Death toll in july revolution reaches 1,423, over 22,000 injured,

    RTV Online, “Death toll in july revolution reaches 1,423, over 22,000 injured,” RTV Online, 2024. Available: https://bit.ly/40sIuK5

    work page 2024

  5. [5]

    Gen z’s leadership strategy in the student movement: A factor analysis on july revolution in bangladesh,

    M. Zahan, A. R. Khan, and G. Mathbor, “Gen z’s leadership strategy in the student movement: A factor analysis on july revolution in bangladesh,” Available at SSRN 5042587 , 2024

    work page 2024

  6. [6]

    July revolution graffiti at dhaka city

    D. Akram, “July revolution graffiti at dhaka city.” https://bit.ly/3Sn0VuU, 2024. Licensed under Creative Commons Attribution 4.0 International (CC BY 4.0)

    work page 2024

  7. [7]

    Graffiti in bangladesh after july revolution

    M. Das, “Graffiti in bangladesh after july revolution.” Available at: https://bit.ly/4d1D2Tf, note = Licensed under Creative Commons CC0 1.0 Public Domain Dedication, 2024

    work page 2024

  8. [8]

    Bangla natural language processing: A comprehensive analysis of classical, machine learning, and deep learning-based methods,

    O. Sen, M. Fuad, M. N. Islam, J. Rabbi, M. Masud, M. K. Hasan, M. A. Awal, A. A. Fime, M. T. H. Fuad, D. Sikder, et al. , “Bangla natural language processing: A comprehensive analysis of classical, machine learning, and deep learning-based methods,” IEEE Access , vol. 10, pp. 38999–39044, 2022

    work page 2022

Show all 37 references
  1. [9]

    Exploring transformer models in the sentiment analysis task for the under-resource bengali language,

    M. N. Hoque, U. Salma, M. J. Uddin, M. M. Ahamad, and S. Aktar, “Exploring transformer models in the sentiment analysis task for the under-resource bengali language,” Natural Language Processing Journal, vol. 8, p. 100091, 2024

    2024

  2. [10]

    Sentiment analysis of bangla language using a new comprehensive dataset bangdsa and the novel feature metric skipbangla-bert,

    M. S. Islam and K. M. Alam, “Sentiment analysis of bangla language using a new comprehensive dataset bangdsa and the novel feature metric skipbangla-bert,” Natural Language Processing Journal, vol. 7, p. 100069, 2024

    2024

  3. [11]

    An- alyzing abusive bangla comments on social media: Nlp & explainable ai,

    K. Ashraf, M. H. Hosen, S. Asgar, M. T. Islam, and S. Nawar, “An- alyzing abusive bangla comments on social media: Nlp & explainable ai,” in 2024 International Conference on Advances in Computing, Communication, Electrical, and Smart Systems (iCACCESS) , pp. 1– 6, 2024

    2024

  4. [12]

    Hybrid cnn lstm approach for sentiment analysis of bengali language comment on facebook,

    S. N. Nobel, S. M. M. R. Swapno, R. A C, H. H. Shajeeb, M. B. Islam, and R. Haque, “Hybrid cnn lstm approach for sentiment analysis of bengali language comment on facebook,” in 2024 International Con- ference on Integrated Circuits and Communication Systems (ICICACS), pp. 1–8, 2024

    2024

  5. [13]

    Sentiment analysis of bangladeshi digital newspaper by using machine learning and natural language processing,

    T. Samin, N. A. Mouri, F. Haque, and M. S. Islam, “Sentiment analysis of bangladeshi digital newspaper by using machine learning and natural language processing,” in 2024 6th International Conference on Electrical Engineering and Information & Communication Technology (ICEEICT...

    2024

  6. [14]

    Bangla e-commerce sentiment analysis optimization using tokeniza- tion and tf-idf,

    S. Sunny, S. Pinky, S. Jalal, M. Kayser, M. Wadud, and N. Mansoor, “Bangla e-commerce sentiment analysis optimization using tokeniza- tion and tf-idf,” in 2024 International Conference on Advances in Com- puting, Communication, Electrical, and Smart Systems (iCACCESS) , pp. 1–...

    2024

  7. [15]

    Bangla sentiment analysis on highly imbalanced data using hybrid cnn-lstm & bangla bert,

    F. Khanam, A. Chakraborty, M. A. Habib, and M. S. Iqbal, “Bangla sentiment analysis on highly imbalanced data using hybrid cnn-lstm & bangla bert,” in 2024 3rd International Conference on Advancement in Electrical and Electronic Engineering (ICAEEE), pp. 1–5, IEEE, 2024

    2024

  8. [16]

    Bert: Pre-training of deep bidirectional transformers for language understanding,

    J. Devlin, “Bert: Pre-training of deep bidirectional transformers for language understanding,” arXiv preprint arXiv:1810.04805 , 2018

    2018 arXiv

  9. [17]

    Unsupervised cross-lingual representation learning at scale,

    A. Conneau, “Unsupervised cross-lingual representation learning at scale,” arXiv preprint arXiv:1911.02116 , 2019

    1911 arXiv

  10. [18]

    Banglabert: Bengali mask language model for bengali language understading,

    S. Sarker, “Banglabert: Bengali mask language model for bengali language understading,” 2020

    2020

  11. [19]

    Visualizing data using t-sne.,

    L. Van der Maaten and G. Hinton, “Visualizing data using t-sne.,” Journal of machine learning research , vol. 9, no. 11, 2008

    2008

  12. [20]

    Umap: Uniform manifold approximation and projection,

    L. McInnes, J. Healy, N. Saul, and L. Großberger, “Umap: Uniform manifold approximation and projection,” Journal of Open Source Software, vol. 3, no. 29, p. 861, 2018

    2018

  13. [21]

    Principal component analysis,

    H. Abdi and L. J. Williams, “Principal component analysis,” Wiley in- terdisciplinary reviews: computational statistics, vol. 2, no. 4, pp. 433– 459, 2010

    2010

  14. [22]

    D. W. Hosmer Jr, S. Lemeshow, and R. X. Sturdivant, Applied logistic regression. John Wiley & Sons, 2013

    2013

  15. [23]

    Support-vector networks,

    C. Cortes, “Support-vector networks,” Machine Learning, 1995

    1995

  16. [24]

    A hybrid machine learning approach utilizing cnn feature extraction with traditional classifier to identify strawberry leaf diseases,

    M. S. Hossen, P. Shaha, and M. Saiduzzaman, “A hybrid machine learning approach utilizing cnn feature extraction with traditional classifier to identify strawberry leaf diseases,” in 4th International Conference on Electrical, Computer and Communication Engineering (ECCE), IEEE, 2025

    2025

  17. [25]

    Induction of decision trees,

    J. R. Quinlan, “Induction of decision trees,” Machine learning, vol. 1, pp. 81–106, 1986

    1986

  18. [26]

    Random forests,

    L. Breiman, “Random forests,” Machine learning , vol. 45, pp. 5–32, 2001

    2001

  19. [27]

    K-nearest neighbor,

    L. E. Peterson, “K-nearest neighbor,” Scholarpedia, vol. 4, no. 2, p. 1883, 2009

    2009

  20. [28]

    An explainable ai-enhanced machine learning approach for cardiovascular disease detection and risk assessment,

    M. E. A. Sourov, M. S. Hossen, P. Shaha, M. M. Hossain, and M. S. Iqbal, “An explainable ai-enhanced machine learning approach for cardiovascular disease detection and risk assessment,” in Proceedings of the International Conference on Quantum Photonics, Artificial Intelligenc...

    2025

  21. [29]

    Xgboost: A scalable tree boosting system,

    T. Chen and C. Guestrin, “Xgboost: A scalable tree boosting system,” in Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining , pp. 785–794, 2016

    2016

  22. [30]

    An explainable ai driven machine learning approach for maternal health risk analysis,

    M. S. Hossen, P. Shaha, M. Saiduzzaman, M. Shovon, A. K. Akhi, and M. S. Iqbal, “An explainable ai driven machine learning approach for maternal health risk analysis,” in 27th International Conference on Computer and Information Technology (ICCIT) , IEEE, 2024

    2024

  23. [31]

    Lightgbm: A highly efficient gradient boosting decision tree,

    G. Ke, Q. Meng, T. Finley, T. Wang, W. Chen, W. Ma, Q. Ye, and T.- Y . Liu, “Lightgbm: A highly efficient gradient boosting decision tree,” Advances in neural information processing systems , vol. 30, 2017

    2017

  24. [32]

    Catboost: unbiased boosting with categorical features,

    L. Prokhorenkova, G. Gusev, A. V orobev, A. V . Dorogush, and A. Gulin, “Catboost: unbiased boosting with categorical features,” Advances in neural information processing systems , vol. 31, 2018

    2018

  25. [33]

    Support vector machine ensemble with bagging,

    H.-C. Kim, S. Pang, H.-M. Je, D. Kim, and S.-Y . Bang, “Support vector machine ensemble with bagging,” in Pattern Recognition with Support Vector Machines: First International Workshop, SVM 2002 Niagara Falls, Canada, August 10, 2002 Proceedings , pp. 397–408, Springer, 2002

    2002

  26. [34]

    An efficient deep learning framework for brain stroke diagnosis using computed tomography (ct) images,

    M. S. Hossen, E. A. Shuvo, S. A. Arif, P. Shaha, M. Saiduzzaman, and M. K. Nasir, “An efficient deep learning framework for brain stroke diagnosis using computed tomography (ct) images,” 2025. arXiv, cs.CV , Available at: https://arxiv.org/abs/2507.03558

    2025

  27. [35]

    A weighted voting classifier based on differential evolution,

    Y . Zhang, H. Zhang, J. Cai, and B. Yang, “A weighted voting classifier based on differential evolution,” in Abstract and applied analysis , vol. 2014, p. 376950, Wiley Online Library, 2014

    2014

  28. [36]

    Greedy function approximation: a gradient boosting machine,

    J. H. Friedman, “Greedy function approximation: a gradient boosting machine,” Annals of statistics , pp. 1189–1232, 2001

    2001

  29. [37]

    Jellyfish species identification: A cnn-based artificial neural network approach,

    M. S. Hossen, M. Saiduzzaman, P. Shaha, and M. K. Nasir, “Jellyfish species identification: A cnn-based artificial neural network approach,” in Proceedings of the International Conference on Quantum Photonics, Artificial Intelligence, and Networking (QPAIN) , IEEE, 2025

    2025

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.