Ghulam Mustafa
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Ghulam Mustafa has a total of 45 co-authors. Here are the top co-authors by number of shared publications:
- Muhammad Tanvir Afzal
Faculty of Computing, Shifa Tameer-e-Millat University, Islamabad 44000, Pakistan
9 shared publications - Isabel de la Torre Díez
Health and Telemedicine Group, University of Valladolid, 47011 Valladolid, Spain
5 shared publications - Ahmad Sami Al-Shamayleh
Department of Data Science and Artificial Intelligence, Faculty of Information Technology, Al-Ahliyya Amman University, Amman 19111, Jordan
4 shared publications - Muhammad Tanvir Afzal
Department of Computing, Shifa Tameer-e-Millat University, Islamabad, Pakistan
4 shared publications - Amer Zaheer
Department of Computer Science, Shifa Tameer-e-Millat University, Islamabad, Pakistan
3 shared publications - Yasir Noman Khalid
Department of Computing, HITEC, Taxila, Pakistan
3 shared publications
Latest Publications
- Open Access
- 11 November 2026
- Published by MDPI AG in Computation
- Vol. 14 (11) , 285
Mining Scholarly Impact: A Career-Adaptive Higher-Order Aggregation Framework for Multi-Domain Researcher Classification
Assessing scholarly contribution across disciplines is challenging because bibliometric indicators capture different aspects of research activity and citation practices vary by field. This study evaluates a weighted cubic-mean composite (Hr) and a career-adaptive extension (DHr) across civil engineering, computer science, mathematics, and neuroscience. The analysis considers 4479 researchers and uses recognition by professional societies as an observable benchmark, while recognizing that awards are imperfect proxies for research quality. reduce selection optimism, we performed nested cross-validation with researcher-level grouping, selecting candidate indicators and metric pairs within inner folds and evaluating the fixed DHr reference formulation on held-out outer folds. On out-of-fold predictions, Hr AUCs were 0.556, 0.692, 0.768, and 0.713, while DHr AUCs were 0.559, 0.690, 0.767, and 0.712 across the four domains, respectively. Paired bootstrap confidence intervals for the Hr–DHr AUC differences included zero in every domain; therefore, these results do not establish that career-adaptive weighting improves discrimination over static aggregation. screening and metric-pair selection were conducted separately in each discipline, these AUCs reflect within-domain validation across four datasets and do not demonstrate transfer to an unseen discipline. The findings support further investigation of composite bibliometric indicators, but independent validation, transparent benchmarking, and fairness assessment are needed before consequential use.
- 11 November 2026
- Published by MDPI AG in Computation
- Vol. 14 (11) , 285
Open AccessMining Scholarly Impact: A Career-Adaptive Higher-Order Aggregation Framework for Multi-Domain Researcher Classification
Assessing scholarly contribution across disciplines is challenging because bibliometric indicators capture different aspects of research activity and citation practices vary by field. This study evaluates a weighted cubic-mean composite (Hr) and a career-adaptive extension (DHr) across civil engineering, computer science, mathematics, and neuroscience. The analysis considers 4479 researchers and uses recognition by professional societies as an observable benchmark, while recognizing that awards are imperfect proxies for research quality. reduce selection optimism, we performed nested cross-validation with researcher-level grouping, selecting candidate indicators and metric pairs within inner folds and evaluating the fixed DHr reference formulation on held-out outer folds. On out-of-fold predictions, Hr AUCs were 0.556, 0.692, 0.768, and 0.713, while DHr AUCs were 0.559, 0.690, 0.767, and 0.712 across the four domains, respectively. Paired bootstrap confidence intervals for the Hr–DHr AUC differences included zero in every domain; therefore, these results do not establish that career-adaptive weighting improves discrimination over static aggregation. screening and metric-pair selection were conducted separately in each discipline, these AUCs reflect within-domain validation across four datasets and do not demonstrate transfer to an unseen discipline. The findings support further investigation of composite bibliometric indicators, but independent validation, transparent benchmarking, and fairness assessment are needed before consequential use.
- Open Access
- 5 November 2026
- Published by MDPI AG in Computation
- Vol. 14 (11) , 278
Does Bigger Mean Better? A Domain-Aware Comparison of Lexicon-Based, TF-IDF, and Transformer Models for Sentiment Analysis
Sentiment analysis systems range from training-free lexicons to compact supervised models and fine-tuned transformer encoders, but fair comparison requires identical test examples and explicit computational accounting. (1) Background: We reassess three sentiment-classification paradigms under matched binary-label evaluation on IMDb movie reviews and Sentiment140 tweets. (2) Methods: After duplicate and label audits, VADER, TF–IDF with validation-tuned Logistic Regression, and three-epoch GPU-fine-tuned DistilBERT were evaluated on the same 5065 IMDb and 10,000 Sentiment140 held-out records within each corpus. Accuracy, macro-F1, confusion matrices, paired McNemar tests, bootstrap intervals, training time, and inference throughput were computed from saved predictions and execution logs. (3) Results: DistilBERT reached 92.00% accuracy on IMDb and 83.41% on Sentiment140, compared with 91.06% and 78.77% for Logistic Regression and 69.83% and 66.54% for VADER. DistilBERT exceeded Logistic Regression by 0.95 percentage points on IMDb (paired 95% CI 0.08–1.76; ) and by 4.64 points on Sentiment140 (CI 3.85–5.41; ), but required GPU fine-tuning and longer measured training. (4) Conclusions: The relative advantage of a transformer depends on the dataset and available computation. Logistic Regression remains a compact, accurate alternative; results are restricted to these two within-domain configurations and do not demonstrate cross-domain transfer or domain causation.- 5 November 2026
- Published by MDPI AG in Computation
- Vol. 14 (11) , 278
Open AccessDoes Bigger Mean Better? A Domain-Aware Comparison of Lexicon-Based, TF-IDF, and Transformer Models for Sentiment Analysis
Sentiment analysis systems range from training-free lexicons to compact supervised models and fine-tuned transformer encoders, but fair comparison requires identical test examples and explicit computational accounting. (1) Background: We reassess three sentiment-classification paradigms under matched binary-label evaluation on IMDb movie reviews and Sentiment140 tweets. (2) Methods: After duplicate and label audits, VADER, TF–IDF with validation-tuned Logistic Regression, and three-epoch GPU-fine-tuned DistilBERT were evaluated on the same 5065 IMDb and 10,000 Sentiment140 held-out records within each corpus. Accuracy, macro-F1, confusion matrices, paired McNemar tests, bootstrap intervals, training time, and inference throughput were computed from saved predictions and execution logs. (3) Results: DistilBERT reached 92.00% accuracy on IMDb and 83.41% on Sentiment140, compared with 91.06% and 78.77% for Logistic Regression and 69.83% and 66.54% for VADER. DistilBERT exceeded Logistic Regression by 0.95 percentage points on IMDb (paired 95% CI 0.08–1.76; ) and by 4.64 points on Sentiment140 (CI 3.85–5.41; ), but required GPU fine-tuning and longer measured training. (4) Conclusions: The relative advantage of a transformer depends on the dataset and available computation. Logistic Regression remains a compact, accurate alternative; results are restricted to these two within-domain configurations and do not demonstrate cross-domain transfer or domain causation. - Open Access
- 3 November 2026
- Published by MDPI AG in Computation
- Vol. 14 (11) , 272
SPE-ECG: An Ontology-Guided Knowledge Representation Framework for Enhancing LLM-Based Code Generation
Large Language Models (LLMs) can translate natural-language instructions into code, but when the input is unstructured clinical protocol text, the generated code may omit doses, monitoring intervals, or safety thresholds that the text states. Such omissions matter because they can propagate into software that supports clinical monitoring. We propose SPE-ECG (Structured Prompt Engineering for Enhanced Code Generation), a transparent, rule-based pipeline that converts unstructured domain text into a JSON prompt using text cleaning, part-of-speech tagging, bigram matching, and hand-built domain dictionaries, without fine-tuning any model. In a case study of 43 rheumatoid arthritis DMARD monitoring protocols, we compared raw and structured prompts on three LLMs (GPT-3.5, LLaMA 2, and Gemma), scoring the generated Python code using version 3.15.1 with BLEU against the raw-prompt code of the other two models. Mean BLEU rose for all three models (relative gains of 55.1%, 32.5%, and 19.9%); the paired differences were significant for GPT-3.5 and LLaMA 2 but not, after correction for multiple comparisons, for Gemma. These results suggest that lightweight input structuring can change how closely different models’ outputs agree, and that it could serve as a low-cost, auditable preparation step for protocol-to-code tools built on free or open models. BLEU does not measure correctness, no ablation or state-of-the-art baseline was run, and clinical validation is required before any use.
- 3 November 2026
- Published by MDPI AG in Computation
- Vol. 14 (11) , 272
Open AccessSPE-ECG: An Ontology-Guided Knowledge Representation Framework for Enhancing LLM-Based Code Generation
Large Language Models (LLMs) can translate natural-language instructions into code, but when the input is unstructured clinical protocol text, the generated code may omit doses, monitoring intervals, or safety thresholds that the text states. Such omissions matter because they can propagate into software that supports clinical monitoring. We propose SPE-ECG (Structured Prompt Engineering for Enhanced Code Generation), a transparent, rule-based pipeline that converts unstructured domain text into a JSON prompt using text cleaning, part-of-speech tagging, bigram matching, and hand-built domain dictionaries, without fine-tuning any model. In a case study of 43 rheumatoid arthritis DMARD monitoring protocols, we compared raw and structured prompts on three LLMs (GPT-3.5, LLaMA 2, and Gemma), scoring the generated Python code using version 3.15.1 with BLEU against the raw-prompt code of the other two models. Mean BLEU rose for all three models (relative gains of 55.1%, 32.5%, and 19.9%); the paired differences were significant for GPT-3.5 and LLaMA 2 but not, after correction for multiple comparisons, for Gemma. These results suggest that lightweight input structuring can change how closely different models’ outputs agree, and that it could serve as a low-cost, auditable preparation step for protocol-to-code tools built on free or open models. BLEU does not measure correctness, no ablation or state-of-the-art baseline was run, and clinical validation is required before any use.
- Open Access
- 14 August 2026
- Published by Springer Nature in Social Network Analysis and Mining
An ontology-guided knowledge graph and retrieval-augmented LLM framework for interpretable mental health assessment
Mental health detection from social media has gained increasing attention due to the spontaneous expression of psychological states on online platforms. However, accurate interpretation remains challenging because social media language is highly contextual, ambiguous, and often includes sarcasm, metaphors, or non-clinical expressions, which can lead to unreliable predictions. Existing approaches based on machine learning, deep learning, transformers, and large language models (LLMs) primarily rely on text-driven patterns and often lack explicit ontology-guided grounding, structured knowledge validation, and interpretability. Although LLMs improve contextual reasoning, their predictions can be inconsistent and prone to hallucination when not supported by structured domain knowledge and a unified cross-component validation mechanism. To address these limitations, this paper proposes an ontology-guided framework that integrates a Knowledge Graph (KG), an Ontology-Informed Retrieval Classifier (ORC), and a Large Language Model for interpretable mental health detection from social media text. The framework leverages the Human Phenotype Ontology (HPO) to guide symptom extraction and Knowledge Graph construction, normalizing informal linguistic expressions into ontology-grounded representations. A key design contribution is an explicit KG–ORC cross-validation consistency gate, which requires both the Knowledge Graph and the ORC module to independently reach agreement before a prediction is committed reducing erroneous inferences from either component alone. The LLM is deliberately restricted to a post-validation role, generating human-readable explanations and supportive recommendations only from validated outputs, thereby reducing hallucination risks and preserving knowledge-consistency. The proposed framework is evaluated on the Dreaddit dataset, a large-scale, multi-domain Reddit corpus for stress and mental health analysis. Empirical evaluation on the HPO guided symptom extraction confirms differential detection rates of 56.04% in stress positive posts versus 31.11% in stress negative posts, validating the signal relevance of the ontology-driven extraction. End-to-end binary classification on the held-out test split achieves an accuracy of 74.82% with an F1-score of 75.64%. Crucially, on the 66.55% of instances where the KG and ORC modules reach consensus, the consistent subset accuracy rises to 78.84% (F1 = 78.28%), compared to 63.03% for the Knowledge Graph alone. The KG–ORC agreement rate across the full test set is 82.02%. These results demonstrate that the KG–ORC cross-validation gate measurably improves predictive reliability over single component baselines, and that ontology-guided, knowledge-grounded reasoning offers a principled path toward interpretable and knowledge-consistent mental health analysis from social media.
- 14 August 2026
- Published by Springer Nature in Social Network Analysis and Mining
Open AccessAn ontology-guided knowledge graph and retrieval-augmented LLM framework for interpretable mental health assessment
Mental health detection from social media has gained increasing attention due to the spontaneous expression of psychological states on online platforms. However, accurate interpretation remains challenging because social media language is highly contextual, ambiguous, and often includes sarcasm, metaphors, or non-clinical expressions, which can lead to unreliable predictions. Existing approaches based on machine learning, deep learning, transformers, and large language models (LLMs) primarily rely on text-driven patterns and often lack explicit ontology-guided grounding, structured knowledge validation, and interpretability. Although LLMs improve contextual reasoning, their predictions can be inconsistent and prone to hallucination when not supported by structured domain knowledge and a unified cross-component validation mechanism. To address these limitations, this paper proposes an ontology-guided framework that integrates a Knowledge Graph (KG), an Ontology-Informed Retrieval Classifier (ORC), and a Large Language Model for interpretable mental health detection from social media text. The framework leverages the Human Phenotype Ontology (HPO) to guide symptom extraction and Knowledge Graph construction, normalizing informal linguistic expressions into ontology-grounded representations. A key design contribution is an explicit KG–ORC cross-validation consistency gate, which requires both the Knowledge Graph and the ORC module to independently reach agreement before a prediction is committed reducing erroneous inferences from either component alone. The LLM is deliberately restricted to a post-validation role, generating human-readable explanations and supportive recommendations only from validated outputs, thereby reducing hallucination risks and preserving knowledge-consistency. The proposed framework is evaluated on the Dreaddit dataset, a large-scale, multi-domain Reddit corpus for stress and mental health analysis. Empirical evaluation on the HPO guided symptom extraction confirms differential detection rates of 56.04% in stress positive posts versus 31.11% in stress negative posts, validating the signal relevance of the ontology-driven extraction. End-to-end binary classification on the held-out test split achieves an accuracy of 74.82% with an F1-score of 75.64%. Crucially, on the 66.55% of instances where the KG and ORC modules reach consensus, the consistent subset accuracy rises to 78.84% (F1 = 78.28%), compared to 63.03% for the Knowledge Graph alone. The KG–ORC agreement rate across the full test set is 82.02%. These results demonstrate that the KG–ORC cross-validation gate measurably improves predictive reliability over single component baselines, and that ontology-guided, knowledge-grounded reasoning offers a principled path toward interpretable and knowledge-consistent mental health analysis from social media.
- 12 August 2026
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 82 (12) , 630
A resource and deadline-aware dynamic load balancing scheduler for cloud jobs
- 12 August 2026
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 82 (12) , 630
A resource and deadline-aware dynamic load balancing scheduler for cloud jobs
- 1 August 2026
- Published by Springer Nature in International Journal of Data Science and Analytics
- Vol. 22 (1) , 252
A data-driven composite index for evaluating research impact: evidence from neuroscience researchers
- 1 August 2026
- Published by Springer Nature in International Journal of Data Science and Analytics
- Vol. 22 (1) , 252
A data-driven composite index for evaluating research impact: evidence from neuroscience researchers
- Open Access
- 22 May 2026
- Published by MDPI AG in Information
- Vol. 17 (6) , 515
Predicting Academic Award Recognition Across Disciplines Using Publication-Based Bibliometric Indices and SHAP-Driven Explainability
Researcher evaluation underpins critical academic decisions, yet traditional bibliometric indicators lack predictive capability and cross-domain generalizability, while most predictive approaches offer limited interpretability and narrow domain validation. This study proposes a SHAP interpretable, multi-domain supervised learning framework for predicting academic award recognition using thirty two publication count-based bibliometric indices. A balanced dataset was constructed across four disciplines, namely Computer Science, Neuroscience, Mathematics, and Civil Engineering, comprising verified awardees from recognized professional societies and matched non-awardee researchers. Eight classifiers were evaluated under stratified five fold cross validation, assessed via accuracy, precision, recall, F1-score, and ROC AUC. The framework achieved domain-specific F1-scores of 0.70 in Computer Science, 0.73 in Neuroscience, 0.72 in Civil Engineering, and 0.78 in Mathematics, with SVM and XGBoost demonstrating the strongest cross-domain robustness across disciplines. SHAP analysis consistently identified normalized h index, h2 family, q2 index, and g index as dominant cross-domain predictors, while domain-specific indicators, including Rm and w indices in Neuroscience and P index in Civil Engineering, reflected disciplinary recognition patterns. By unifying publication-based feature engineering, multi-domain classification, and SHAP explainability within a single reproducible pipeline, this framework offers a scalable, transparent, and evidence-based tool for institutional researcher evaluation.
- 22 May 2026
- Published by MDPI AG in Information
- Vol. 17 (6) , 515
Open AccessPredicting Academic Award Recognition Across Disciplines Using Publication-Based Bibliometric Indices and SHAP-Driven Explainability
Researcher evaluation underpins critical academic decisions, yet traditional bibliometric indicators lack predictive capability and cross-domain generalizability, while most predictive approaches offer limited interpretability and narrow domain validation. This study proposes a SHAP interpretable, multi-domain supervised learning framework for predicting academic award recognition using thirty two publication count-based bibliometric indices. A balanced dataset was constructed across four disciplines, namely Computer Science, Neuroscience, Mathematics, and Civil Engineering, comprising verified awardees from recognized professional societies and matched non-awardee researchers. Eight classifiers were evaluated under stratified five fold cross validation, assessed via accuracy, precision, recall, F1-score, and ROC AUC. The framework achieved domain-specific F1-scores of 0.70 in Computer Science, 0.73 in Neuroscience, 0.72 in Civil Engineering, and 0.78 in Mathematics, with SVM and XGBoost demonstrating the strongest cross-domain robustness across disciplines. SHAP analysis consistently identified normalized h index, h2 family, q2 index, and g index as dominant cross-domain predictors, while domain-specific indicators, including Rm and w indices in Neuroscience and P index in Civil Engineering, reflected disciplinary recognition patterns. By unifying publication-based feature engineering, multi-domain classification, and SHAP explainability within a single reproducible pipeline, this framework offers a scalable, transparent, and evidence-based tool for institutional researcher evaluation.
- Open Access
- 20 May 2026
- Published by MDPI AG in Computation
- Vol. 14 (5) , 115
Quantifying Domain-Specific Risk Signals in Lung Cancer Severity Prediction: A Multi-Domain Ablation Study Using XGBoost and SHAP
Predictive modeling for lung cancer severity often struggles with the high dimensionality and multi-domain nature of risk factors. While individual contributors like smoking are well-documented, the relative predictive weight of lifestyle, environmental, and genetic domains remains insufficiently quantified in integrated frameworks. This study proposes an explainable machine learning approach using an XGBoost classifier to evaluate these three distinct risk domains. Utilizing the UCI Machine Learning Repository Lung Cancer Dataset, we implemented a domain-wise ablation study to isolate the predictive signal of each factor group. To ensure scientific rigor and address the “black box” nature of ensemble models, we employed 5-fold stratified cross-validation and SHAP (Shapley Additive Explanations) for feature-level transparency. Our results demonstrate that the integrated model achieves a classification accuracy of 95.7% (AUC-ROC = 0.98) on this dataset. Notably, ablation analysis revealed that the Lifestyle domain retained the highest standalone predictive performance (92.9%), followed by the Genetic/Clinical domain (94.6%), while the Environmental domain showed a more pronounced performance drop (73.3%), suggesting differential information density across risk categories. SHAP analysis identified cumulative smoking exposure as the primary feature influencing model predictions within this dataset. This study presents a proof-of-concept interpretable framework for lung cancer risk stratification, demonstrating that domain-wise ablation combined with explainable AI can provide transparent, feature-level insight to support rather than replace clinical judgment in settings where comprehensive diagnostic testing may be limited.
- 20 May 2026
- Published by MDPI AG in Computation
- Vol. 14 (5) , 115
Open AccessQuantifying Domain-Specific Risk Signals in Lung Cancer Severity Prediction: A Multi-Domain Ablation Study Using XGBoost and SHAP
Predictive modeling for lung cancer severity often struggles with the high dimensionality and multi-domain nature of risk factors. While individual contributors like smoking are well-documented, the relative predictive weight of lifestyle, environmental, and genetic domains remains insufficiently quantified in integrated frameworks. This study proposes an explainable machine learning approach using an XGBoost classifier to evaluate these three distinct risk domains. Utilizing the UCI Machine Learning Repository Lung Cancer Dataset, we implemented a domain-wise ablation study to isolate the predictive signal of each factor group. To ensure scientific rigor and address the “black box” nature of ensemble models, we employed 5-fold stratified cross-validation and SHAP (Shapley Additive Explanations) for feature-level transparency. Our results demonstrate that the integrated model achieves a classification accuracy of 95.7% (AUC-ROC = 0.98) on this dataset. Notably, ablation analysis revealed that the Lifestyle domain retained the highest standalone predictive performance (92.9%), followed by the Genetic/Clinical domain (94.6%), while the Environmental domain showed a more pronounced performance drop (73.3%), suggesting differential information density across risk categories. SHAP analysis identified cumulative smoking exposure as the primary feature influencing model predictions within this dataset. This study presents a proof-of-concept interpretable framework for lung cancer risk stratification, demonstrating that domain-wise ablation combined with explainable AI can provide transparent, feature-level insight to support rather than replace clinical judgment in settings where comprehensive diagnostic testing may be limited.
- 18 April 2026
- Published by Springer Nature in Scientometrics
- Vol. 131 (5) , 2885-2925
A hybrid framework for predicting scholarly recognition of authors using network centrality and generative AI embeddings
- 18 April 2026
- Published by Springer Nature in Scientometrics
- Vol. 131 (5) , 2885-2925
A hybrid framework for predicting scholarly recognition of authors using network centrality and generative AI embeddings
- Open Access
- 2 April 2026
- Published by Springer Nature in Journal of Big Data
- Vol. 13 (1)
Venue-aware researcher impact assessment using genetic programming
Researcher evaluation remains a central challenge in scientometrics, where reliable, transparent, and context-sensitive methods are required for decisions related to recruitment, promotion, and awards. Traditional assessment approaches rely heavily on bibliometric indices such as the h -index and its variants; however, these measures often neglect publication venue quality and provide limited discriminatory capability. This study proposes an integrated framework that combines retrospective bibliometric analysis, venue-aware modeling, and Genetic Programming (GP)-based symbolic regression for interpretable researcher evaluation. A balanced dataset of 1,200 computer science authors, equally divided between awardees and non-awardees, was constructed using Google Scholar and Publish or Perish . First, sixty-four bibliometric indices were computed to establish a retrospective ranking baseline, where the $$h_2$$ upper-index achieved the strongest performance by identifying 73% of awardees within top-ranked positions. Second, a venue-aware dataset incorporating journals, conferences, books, and patents was developed, and neural-network–estimated venue contributions indicated journals as the dominant factor (84%). Finally, GP-based symbolic regression was applied to evolve interpretable closed-form equations integrating venue-level features. The best GP-derived model identified 91% of awardees, outperforming both the retrospective baseline and the venue-aware linear model while maintaining interpretability. These findings demonstrate that combining venue-aware modeling with interpretable evolutionary computation provides a more accurate, transparent, and equitable framework for researcher evaluation, with practical implications for academic institutions and research policy design.- 2 April 2026
- Published by Springer Nature in Journal of Big Data
- Vol. 13 (1)
Open AccessVenue-aware researcher impact assessment using genetic programming
Researcher evaluation remains a central challenge in scientometrics, where reliable, transparent, and context-sensitive methods are required for decisions related to recruitment, promotion, and awards. Traditional assessment approaches rely heavily on bibliometric indices such as the h -index and its variants; however, these measures often neglect publication venue quality and provide limited discriminatory capability. This study proposes an integrated framework that combines retrospective bibliometric analysis, venue-aware modeling, and Genetic Programming (GP)-based symbolic regression for interpretable researcher evaluation. A balanced dataset of 1,200 computer science authors, equally divided between awardees and non-awardees, was constructed using Google Scholar and Publish or Perish . First, sixty-four bibliometric indices were computed to establish a retrospective ranking baseline, where the $$h_2$$ upper-index achieved the strongest performance by identifying 73% of awardees within top-ranked positions. Second, a venue-aware dataset incorporating journals, conferences, books, and patents was developed, and neural-network–estimated venue contributions indicated journals as the dominant factor (84%). Finally, GP-based symbolic regression was applied to evolve interpretable closed-form equations integrating venue-level features. The best GP-derived model identified 91% of awardees, outperforming both the retrospective baseline and the venue-aware linear model while maintaining interpretability. These findings demonstrate that combining venue-aware modeling with interpretable evolutionary computation provides a more accurate, transparent, and equitable framework for researcher evaluation, with practical implications for academic institutions and research policy design.