pith. sign in

arxiv: 2606.03864 · v1 · pith:WJ5VTGJZnew · submitted 2026-06-02 · 💻 cs.SI · cs.CY· cs.DL· cs.LG· physics.soc-ph

Explainable Forecasting of Scientific Breakthroughs from Concept Network Dynamics

Thomas Maillart , Thibaut Chataing , Ntorina Antoni , David Dosu , Paul Bagourd , Julian Jang-Jaccard , Alain Mermoud This is my paper

Pith reviewed 2026-06-28 07:49 UTC · model grok-4.3

classification 💻 cs.SI cs.CYcs.DLcs.LGphysics.soc-ph
keywords concept networkslink predictionscientific breakthroughsmachine learningOpenAlexexplainable forecastingnetwork dynamicsresearch strategy
0
0 comments X

The pith

A two-stage LightGBM model using 59 network features forecasts the formation and future weight of concept pairs that precede scientific breakthroughs with ROC-AUC above 0.95.

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

The paper develops a machine-learning system that predicts when new connections between scientific concepts will form and how strongly they will grow by tracking the evolution of large concept networks. It extracts 59 structural and semantic measures from publication data and feeds them into a classifier followed by a regressor that estimates link intensity. The approach improves accuracy over earlier link-prediction methods while ensuring every forecast can be traced to specific, auditable network properties rather than hidden patterns. If the predictions hold across domains, they could surface promising research recombinations earlier than review processes that rely solely on expert intuition. The work validates the system on technology and biomedical fields and shows alignment with known cases in quantum technologies.

Core claim

The authors claim that the evolution of concept co-occurrence networks contains detectable structural and semantic signals that a two-stage LightGBM classifier-regressor can use to jointly forecast link formation and link weight at one- to five-year horizons, attaining ROC-AUC values between 0.954 and 0.967 in multiple domains while remaining fully explainable through feature importance.

What carries the argument

The two-stage LightGBM model operating on 59 semantic and topological features extracted from OpenAlex concept networks, where the first stage classifies future link existence and the second stage regresses expected link weight.

If this is right

  • Forecast accuracy exceeds prior models' roughly 0.90 AUC without any domain-specific re-tuning.
  • Structural features, especially Adamic-Adar similarity and degree-based Hadamard measures, consistently rank as the most important predictors.
  • Every prediction can be audited because it depends on explicit network measures rather than black-box embeddings.
  • Case studies in quantum annealing and AI-enabled quantum architectures align model outputs with expert-identified technological convergences.
  • The forecasts support a three-layer decision process of detection, expert translation, and institutional integration for research strategy.

Where Pith is reading between the lines

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

  • If the features generalize, the same pipeline could be applied to patent citation networks to forecast technological recombinations.
  • The finding that breakthroughs emerge in tightly connected sub-networks could be tested by checking citation impact of high-prediction pairs in dense regions.
  • Embedding the model in funding pipelines might shorten the lag from signal to resource allocation, though this requires separate validation of policy effects.
  • Adding temporal author-collaboration features could test whether social structure modulates the predictive power of pure concept networks.

Load-bearing premise

The assumption that the structural and semantic features extracted from OpenAlex concept networks are sufficient and generalizable predictors of breakthrough-relevant recombinations across domains without domain-specific tuning.

What would settle it

Apply the trained model without re-tuning to a new scientific domain or later time window and measure whether ROC-AUC falls below 0.90 or whether pairs predicted to form strong links fail to produce highly cited papers within the forecast horizon.

Figures

Figures reproduced from arXiv: 2606.03864 by Alain Mermoud, David Dosu, Julian Jang-Jaccard, Ntorina Antoni, Paul Bagourd, Thibaut Chataing, Thomas Maillart.

Figure 1
Figure 1. Figure 1: Various possibilities for structural change in the concept graph. Breakthroughs can appear as new nodes, new connected pairs, triplets, quadruplets, etc. or even more specific structures. Here, we consider the formation and the evolution of pairs to gauge for breakthroughs, as they are simple enough and appear sufficiently often in the graph to allow efficient machine learning. Here, two assumptions are ce… view at source ↗
Figure 2
Figure 2. Figure 2: A. Evidence of proportional growth in edge weights. B. Evidence of significant hazard rate [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Machine learning pipeline: A. The input is the evolution of the concept graph. B. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: (Upper panel) Performance of the link prediction model. A. Accuracy decreases only [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Feature importance (gain) for prediction horizons [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Use-case validation. A: Quantum Annealing shows predicted reinforcement of core physics and optimization concepts. B: AI-accelerated quantum computing shows predicted strengthening of interdisciplinary clusters. the Adamic–Adar and degree-Hadamard mechanisms identified in Section 5 are not artefacts of a single testbed, while weight-intensity forecasts should be interpreted jointly with a domain’s growth v… view at source ↗
Figure 7
Figure 7. Figure 7: ROC–AUC versus prediction horizon across four research domains. All domains remain [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: RMSLE versus prediction horizon. Steady-growth domains (advanced materials, neuro [PITH_FULL_IMAGE:figures/full_fig_p011_8.png] view at source ↗
read the original abstract

We introduce an explainable machine-learning approach that forecasts the structural precursors of scientific breakthroughs -- the emergence and intensification of links between research concepts -- by modelling how OpenAlex concept networks evolve over time. Using 59 semantic and topological features, a two-stage LightGBM model jointly predicts the formation and the future weight of concept pairs, adding a regression stage that quantifies expected intensity to prior link-existence forecasts. Relative to the state of the art, the approach improves accuracy and explainability at once: comparative validation across four technology and biomedical domains yields ROC-AUC in [0.954, 0.967] at all horizons without re-tuning, exceeding the roughly 0.90 of prior models, while every forecast rests on structural, auditable features rather than opaque embeddings. Classification performance is high (AUC about 0.95) and regression remains stable (RMSLE 0.45 to 0.6 over one to five years). Feature attribution shows that structural factors -- particularly Adamic-Adar similarity and degree-based Hadamard measures -- consistently drive accuracy, suggesting that breakthrough-relevant recombinations emerge in tightly connected sub-networks. Two expert-anchored cases, quantum annealing and AI-enabled quantum architectures, show the model surfacing technological convergence consistent with expert expectations. We then outline a three-layer decision architecture -- detection, expert translation, institutional integration -- that turns these forecasts into evidence-based research strategy and policy, anchored in open data and explainable features.

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.

Referee Report

1 major / 2 minor

Summary. The paper introduces a two-stage LightGBM model that uses 59 semantic and topological features extracted from OpenAlex concept networks to jointly predict the formation of new links between concepts and their future weights. This is framed as forecasting the structural precursors of scientific breakthroughs. Across four domains and multiple time horizons, the model achieves ROC-AUC values in [0.954, 0.967] without re-tuning, outperforming prior link-prediction baselines around 0.90, with regression RMSLE between 0.45 and 0.6. Feature attributions highlight structural measures such as Adamic-Adar similarity and degree-based Hadamard products as key drivers. Two qualitative case studies on quantum annealing and AI-enabled quantum architectures are presented, followed by a proposed three-layer decision architecture for research strategy.

Significance. If the link-formation predictions can be shown to correspond to actual breakthrough events, the work would provide a useful, explainable tool for science forecasting that relies on auditable structural features and open data rather than opaque embeddings. The cross-domain stability without hyperparameter retuning and the addition of a regression stage for intensity quantification are concrete strengths. The emphasis on feature attribution also supports interpretability claims.

major comments (1)
  1. [Abstract / evaluation and case-study sections] Abstract and the evaluation/case-study sections: the central claim is that the model forecasts 'structural precursors of scientific breakthroughs' via concept-pair link emergence and intensification. However, all reported metrics (ROC-AUC, RMSLE) evaluate only the auxiliary task of link prediction on historical network evolution. No quantitative validation is supplied showing that high-weight predicted links are enriched among known breakthrough papers, exhibit differential citation impact, or align with external breakthrough indicators. The two expert-anchored cases supply only qualitative consistency. This disconnect is load-bearing for the title, abstract, and policy implications.
minor comments (2)
  1. [Abstract] The abstract states that the model was validated 'across four technology and biomedical domains' but does not name the domains or provide the exact data splits and temporal validation protocol used.
  2. [Methods / feature engineering] Feature definitions and extraction details for the 59 semantic and topological features are referenced but not fully enumerated or justified with respect to potential multicollinearity or domain-specific biases.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the detailed and constructive report. The major comment identifies a substantive gap between the paper's framing and its quantitative evidence; we address it directly below with a proposed partial revision to improve precision without altering the technical contributions.

read point-by-point responses

  1. Referee: [Abstract / evaluation and case-study sections] Abstract and the evaluation/case-study sections: the central claim is that the model forecasts 'structural precursors of scientific breakthroughs' via concept-pair link emergence and intensification. However, all reported metrics (ROC-AUC, RMSLE) evaluate only the auxiliary task of link prediction on historical network evolution. No quantitative validation is supplied showing that high-weight predicted links are enriched among known breakthrough papers, exhibit differential citation impact, or align with external breakthrough indicators. The two expert-anchored cases supply only qualitative consistency. This disconnect is load-bearing for the title, abstract, and policy implications.

    Authors: We agree that the evaluation is confined to link-prediction and regression metrics on historical concept-network evolution, with no quantitative tests for enrichment among known breakthrough papers, differential citation impact, or alignment with external indicators. The two case studies remain qualitative. This is a fair and load-bearing observation. We will revise the abstract to describe the model as forecasting concept-link formation and intensification (positioned as structural precursors), add an explicit limitations paragraph noting the absence of direct quantitative breakthrough validation, and moderate the policy-implications discussion to reflect the current evidence base. These changes constitute a partial revision that clarifies scope while preserving the reported link-prediction results and feature-attribution findings. revision: partial

Circularity Check

0 steps flagged

No significant circularity; standard supervised link-prediction pipeline

full rationale

The paper extracts 59 features from OpenAlex concept networks at prior time slices and trains a two-stage LightGBM classifier-regressor to predict future link formation and weight on temporally held-out data. Reported metrics (ROC-AUC 0.954-0.967, RMSLE 0.45-0.6) are ordinary out-of-sample performance numbers; they do not reduce to the training inputs by construction. No equations equate a derived quantity to a fitted parameter, no self-citation supplies a uniqueness theorem, and the interpretive step that equates link emergence with 'structural precursors of breakthroughs' is an explicit modeling assumption rather than a hidden definitional loop inside the derivation. The pipeline is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the quality of OpenAlex data and the assumption that network features predict future links without additional domain knowledge.

free parameters (1)
  • LightGBM model parameters
    Hyperparameters and weights are fitted during training on historical network data.
axioms (1)
  • domain assumption Concept networks from OpenAlex accurately represent research concept relationships over time.
    The model relies on this data source being representative of scientific activity.

pith-pipeline@v0.9.1-grok · 5828 in / 1285 out tokens · 33390 ms · 2026-06-28T07:49:33.770435+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

192 extracted references · 125 canonical work pages · 4 internal anchors

  1. [1]

    European Commission , year =. 2030

    2030

  2. [2]

    European Commission , year =. Quantum

  3. [3]

    2020 , doi =

    Anticipatory innovation governance:. 2020 , doi =

    2020

  4. [4]

    2026 , doi =

    Quantum. 2026 , doi =

    2026

  5. [5]

    Systems Research and Behavioral Science , author =

    Complexity and the productivity of innovation , volume =. Systems Research and Behavioral Science , author =. 2010 , keywords =. doi:10.1002/sres.1057 , abstract =

    work page doi:10.1002/sres.1057 2010

  6. [6]

    Bernhardt, Chris , month = mar, year =. Quantum

  7. [7]

    Journal of Informetrics , author =

    The memory of science:. Journal of Informetrics , author =. 2018 , keywords =. doi:10.1016/j.joi.2018.06.005 , abstract =

    work page doi:10.1016/j.joi.2018.06.005 2018

  8. [8]

    Journal of data and information science , author =

    Science mapping: a systematic review of the literature , volume =. Journal of data and information science , author =

  9. [9]

    The structure of scientific revolutions , volume =

    Kuhn, Thomas S , year =. The structure of scientific revolutions , volume =

  10. [10]

    Diversity and

    Page, Scott , month = nov, year =. Diversity and. doi:10.1515/9781400835140 , keywords =

    work page doi:10.1515/9781400835140

  11. [11]

    Psychological Review , author =

    Blind variation and selective retentions in creative thought as in other knowledge processes , volume =. Psychological Review , author =. 1960 , note =. doi:10.1037/h0040373 , abstract =

    work page doi:10.1037/h0040373 1960

  12. [12]

    Research Policy , author =

    Scientific novelty and technological impact , volume =. Research Policy , author =. 2019 , keywords =. doi:10.1016/j.respol.2019.01.019 , abstract =

    work page doi:10.1016/j.respol.2019.01.019 2019

  13. [13]

    Journal of The Royal Society Interface , author =

    Invention as a combinatorial process: evidence from. Journal of The Royal Society Interface , author =. 2015 , note =. doi:10.1098/rsif.2015.0272 , abstract =

    work page doi:10.1098/rsif.2015.0272 2015

  14. [14]

    Annual Review of Information Science and Technology (ARIST) , author =

    Visualizing. Annual Review of Information Science and Technology (ARIST) , author =. 2003 , note =

    2003

  15. [15]

    Atlas of

    Börner, Katy , month = sep, year =. Atlas of

  16. [16]

    Nature Communications , author =

    Surprising combinations of research contents and contexts are related to impact and emerge with scientific outsiders from distant disciplines , volume =. Nature Communications , author =. 2023 , note =. doi:10.1038/s41467-023-36741-4 , abstract =

    work page doi:10.1038/s41467-023-36741-4 2023

  17. [17]

    Technological Forecasting and Social Change , author =

    Predicting scientific breakthroughs based on knowledge structure variations , volume =. Technological Forecasting and Social Change , author =. 2021 , keywords =. doi:10.1016/j.techfore.2020.120502 , abstract =

    work page doi:10.1016/j.techfore.2020.120502 2021

  18. [18]

    Nature , author =

    Large teams develop and small teams disrupt science and technology , volume =. Nature , author =. 2019 , note =. doi:10.1038/s41586-019-0941-9 , abstract =

    work page doi:10.1038/s41586-019-0941-9 2019

  19. [19]

    Research Policy , author =

    International research collaboration:. Research Policy , author =. 2019 , keywords =. doi:10.1016/j.respol.2019.01.002 , abstract =

    work page doi:10.1016/j.respol.2019.01.002 2019

  20. [20]

    Embedding technique and network analysis of scientific innovations emergence in an

    Brodiuk, Serhii and Palchykov, Vasyl and Holovatch, Yurij , month = aug, year =. Embedding technique and network analysis of scientific innovations emergence in an. 2020. doi:10.1109/DSMP47368.2020.9204220 , abstract =

    work page doi:10.1109/dsmp47368.2020.9204220 2020

  21. [21]

    Tonn, Bruce , month = nov, year =. The. doi:10.1016/j.futures.2010.08.015 , journal =

    work page doi:10.1016/j.futures.2010.08.015 2010

  22. [22]

    Brian , month = aug, year =

    Arthur, W. Brian , month = aug, year =. The

  23. [23]

    Technological Forecasting and Social Change , author =

    Measuring security development in information technologies:. Technological Forecasting and Social Change , author =. 2023 , pages =. doi:10.1016/j.techfore.2023.122316 , abstract =

    work page doi:10.1016/j.techfore.2023.122316 2023

  24. [24]

    Physica A: Statistical Mechanics and its Applications , author =

    Measuring the preferential attachment mechanism in citation networks , volume =. Physica A: Statistical Mechanics and its Applications , author =. 2008 , keywords =. doi:10.1016/j.physa.2008.03.017 , abstract =

    work page doi:10.1016/j.physa.2008.03.017 2008

  25. [25]

    Nature Machine Intelligence , author =

    From local explanations to global understanding with explainable. Nature Machine Intelligence , author =. 2020 , note =. doi:10.1038/s42256-019-0138-9 , abstract =

    work page doi:10.1038/s42256-019-0138-9 2020

  26. [26]

    Lundberg, Scott M and Lee, Su-In , year =. A. Advances in

  27. [27]

    Advances in

    Ke, Guolin and Meng, Qi and Finley, Thomas and Wang, Taifeng and Chen, Wei and Ma, Weidong and Ye, Qiwei and Liu, Tie-Yan , year =. Advances in

  28. [28]

    Akiba, Takuya and Sano, Shotaro and Yanase, Toshihiko and Ohta, Takeru and Koyama, Masanori , month = jul, year =. Optuna:. Proceedings of the 25th. doi:10.1145/3292500.3330701 , abstract =

    work page doi:10.1145/3292500.3330701

  29. [29]

    Proceedings of the National Academy of Sciences , author =

    The structure of scientific collaboration networks , volume =. Proceedings of the National Academy of Sciences , author =. 2001 , note =. doi:10.1073/pnas.98.2.404 , abstract =

    work page doi:10.1073/pnas.98.2.404 2001

  30. [30]

    Vinod and Jolad, Shivakumar , month = nov, year =

    Enduri, Murali Krishna and Reddy, I. Vinod and Jolad, Shivakumar , month = nov, year =. Does. 2015 11th. doi:10.1109/SITIS.2015.60 , abstract =

    work page doi:10.1109/sitis.2015.60 2015

  31. [31]

    The structure of scientific collaboration networks

  32. [32]

    Proceedings of the National Academy of Sciences , author =

    Choosing experiments to accelerate collective discovery , volume =. Proceedings of the National Academy of Sciences , author =. 2015 , note =. doi:10.1073/pnas.1509757112 , abstract =

    work page doi:10.1073/pnas.1509757112 2015

  33. [33]

    and Osborne, Francesco and Thanapalasingam, Thiviyan and Motta, Enrico , editor =

    Salatino, Angelo A. and Osborne, Francesco and Thanapalasingam, Thiviyan and Motta, Enrico , editor =. The. Digital. 2019 , keywords =. doi:10.1007/978-3-030-30760-8_26 , abstract =

    work page doi:10.1007/978-3-030-30760-8_26 2019

  34. [34]

    PeerJ Computer Science , author =

    How are topics born?. PeerJ Computer Science , author =. 2017 , note =. doi:10.7717/peerj-cs.119 , abstract =

    work page doi:10.7717/peerj-cs.119 2017

  35. [35]

    FUTURES & FORESIGHT SCIENCE , author =

    Horizon. FUTURES & FORESIGHT SCIENCE , author =. 2020 , note =. doi:10.1002/ffo2.23 , abstract =

    work page doi:10.1002/ffo2.23 2020

  36. [36]

    Science and Public Policy , author =

    Facing the future:. Science and Public Policy , author =. 2012 , pages =. doi:10.1093/scipol/scs021 , abstract =

    work page doi:10.1093/scipol/scs021 2012

  37. [37]

    The handbook of technology foresight: concepts and practice , publisher =

    Georghiou, Luke , year =. The handbook of technology foresight: concepts and practice , publisher =

  38. [38]

    Technological Forecasting and Social Change , author =

    Using scenarios for roadmapping:. Technological Forecasting and Social Change , author =. 2010 , keywords =. doi:10.1016/j.techfore.2010.03.003 , abstract =

    work page doi:10.1016/j.techfore.2010.03.003 2010

  39. [39]

    Foresight:

    Cuhls, Kerstin and Dönitz, Ewa and Erdmann, Lorenz and Gransche, Bruno and Kimpeler, Simone and Schirrmeister, Elna and Warnke, Philine , year =. Foresight:. Systems and innovation research in transition:

  40. [40]

    Research Policy , author =

    Three frames for innovation policy:. Research Policy , author =. 2018 , keywords =. doi:10.1016/j.respol.2018.08.011 , abstract =

    work page doi:10.1016/j.respol.2018.08.011 2018

  41. [41]

    Systems and

    Edler, Jakob and Walz, Rainer , month = sep, year =. Systems and

  42. [42]

    Futures , author =

    Exploring the governance and implementation of sustainable development initiatives through blockchain technology , volume =. Futures , author =. 2020 , keywords =. doi:10.1016/j.futures.2020.102611 , abstract =

    work page doi:10.1016/j.futures.2020.102611 2020

  43. [43]

    Using scenarios for roadmapping:

  44. [44]

    Engineering Proceedings , author =

    Towards. Engineering Proceedings , author =. 2022 , note =. doi:10.3390/engproc2022018017 , abstract =

    work page doi:10.3390/engproc2022018017 2022

  45. [45]

    Social Networks , author =

    Friends and neighbors on the. Social Networks , author =. 2003 , keywords =. doi:10.1016/S0378-8733(03)00009-1 , abstract =

    work page doi:10.1016/s0378-8733(03)00009-1 2003

  46. [46]

    and Malevergne, Yannick and Sornette, Didier , month = nov, year =

    Saichev, Alexander I. and Malevergne, Yannick and Sornette, Didier , month = nov, year =. Theory of

  47. [47]

    Physical Review Letters , author =

    Empirical. Physical Review Letters , author =. 2008 , note =. doi:10.1103/PhysRevLett.101.218701 , abstract =

    work page doi:10.1103/physrevlett.101.218701 2008

  48. [48]

    , month = may, year =

    Schumpeter, Joseph A. , month = may, year =. Capitalism,

  49. [49]

    Management Science , author =

    Recombinant. Management Science , author =. 2001 , note =. doi:10.1287/mnsc.47.1.117.10671 , abstract =

    work page doi:10.1287/mnsc.47.1.117.10671 2001

  50. [50]

    Science , author =

    The. Science , author =. 2007 , pmid =. doi:10.1126/science.1136099 , abstract =

    work page doi:10.1126/science.1136099 2007

  51. [51]

    Bradford, Anu , year =. Digital

  52. [52]

    Scientometrics , author =

    Forecasting emerging technologies using data augmentation and deep learning , volume =. Scientometrics , author =. 2020 , keywords =. doi:10.1007/s11192-020-03351-6 , abstract =

    work page doi:10.1007/s11192-020-03351-6 2020

  53. [53]

    Nature Machine Intelligence , author =

    Forecasting the future of artificial intelligence with machine learning-based link prediction in an exponentially growing knowledge network , volume =. Nature Machine Intelligence , author =. 2023 , note =. doi:10.1038/s42256-023-00735-0 , abstract =

    work page doi:10.1038/s42256-023-00735-0 2023

  54. [54]

    Machine Learning: Science and Technology , author =

    Forecasting high-impact research topics via machine learning on evolving knowledge graphs , volume =. Machine Learning: Science and Technology , author =. 2025 , note =. doi:10.1088/2632-2153/add6ef , abstract =

    work page doi:10.1088/2632-2153/add6ef 2025

  55. [55]

    Scientometrics , author =

    Mapping and comparing the technology evolution paths of scientific papers and patents: an integrated approach for forecasting technology trends , volume =. Scientometrics , author =. 2024 , keywords =. doi:10.1007/s11192-024-04961-0 , abstract =

    work page doi:10.1007/s11192-024-04961-0 2024

  56. [56]

    A deep learning-based method for predicting the emerging degree of research topics using emerging index , journal =

    Yang, Zhenyu and Zhang, Wenyu and Wang, Zhimin and Huang, Xiaoling , year =. A deep learning-based method for predicting the emerging degree of research topics using emerging index , journal =

  57. [57]

    Oikos , author =

    Entropy and diversity , volume =. Oikos , author =. 2006 , note =. doi:10.1111/j.2006.0030-1299.14714.x , abstract =

    work page doi:10.1111/j.2006.0030-1299.14714.x 2006

  58. [58]

    Scientometrics , author =

    Measuring destabilization and consolidation in scientific knowledge evolution , volume =. Scientometrics , author =. 2022 , keywords =. doi:10.1007/s11192-022-04479-3 , abstract =

    work page doi:10.1007/s11192-022-04479-3 2022

  59. [59]

    Science , author =

    Quantifying the evolution of individual scientific impact , volume =. Science , author =. 2016 , note =. doi:10.1126/science.aaf5239 , number =

    work page doi:10.1126/science.aaf5239 2016

  60. [60]

    Science , author =

    Atypical. Science , author =. 2013 , note =. doi:10.1126/science.1240474 , abstract =

    work page doi:10.1126/science.1240474 2013

  61. [61]

    https://designthinkingmeite.web.unc.edu/wp-content/uploads/sites/22337/2020/02/

    2020

  62. [62]

    Chen, Chaomei , year =

  63. [63]

    Foresight , author =

    Dynamic foresight evaluation , volume =. Foresight , author =. 2012 , pages =. doi:10.1108/14636681211210378 , abstract =

    work page doi:10.1108/14636681211210378 2012

  64. [64]

    Technological Forecasting and Social Change , author =

    The origins of the concept of ‘foresight’ in science and technology:. Technological Forecasting and Social Change , author =. 2010 , keywords =. doi:10.1016/j.techfore.2010.06.009 , abstract =

    work page doi:10.1016/j.techfore.2010.06.009 2010

  65. [65]

    Scientific prize network predicts who pushes the boundaries of science

  66. [66]

    Science , author =

    Metaknowledge , volume =. Science , author =. 2011 , note =. doi:10.1126/science.1201765 , abstract =

    work page doi:10.1126/science.1201765 2011

  67. [67]

    Reviews of Modern Physics , author =

    Machine learning and the physical sciences , volume =. Reviews of Modern Physics , author =. 2019 , note =. doi:10.1103/RevModPhys.91.045002 , abstract =

    work page doi:10.1103/revmodphys.91.045002 2019

  68. [68]

    Reports on Progress in Physics , author =

    Machine learning & artificial intelligence in the quantum domain: a review of recent progress , volume =. Reports on Progress in Physics , author =. 2018 , note =. doi:10.1088/1361-6633/aab406 , abstract =

    work page doi:10.1088/1361-6633/aab406 2018

  69. [69]

    Proceedings of the National Academy of Sciences , author =

    Predicting research trends with semantic and neural networks with an application in quantum physics , volume =. Proceedings of the National Academy of Sciences , author =. 2020 , note =. doi:10.1073/pnas.1914370116 , abstract =

    work page doi:10.1073/pnas.1914370116 2020

  70. [70]

    Advanced Intelligent Systems , author =

    Forecasting. Advanced Intelligent Systems , author =. doi:10.1002/aisy.202401124 , abstract =

    work page doi:10.1002/aisy.202401124

  71. [71]

    International Journal of Intelligent Systems , author =

    A scientific research topic trend prediction model based on multi-. International Journal of Intelligent Systems , author =. 2022 , note =. doi:10.1002/int.22846 , abstract =

    work page doi:10.1002/int.22846 2022

  72. [72]

    Technological Forecasting and Social Change , author =

    An exploration method for technology forecasting that combines link prediction with graph embedding:. Technological Forecasting and Social Change , author =. 2024 , keywords =. doi:10.1016/j.techfore.2024.123736 , abstract =

    work page doi:10.1016/j.techfore.2024.123736 2024

  73. [73]

    Journal of Informetrics , author =

    Predicting scientific research trends based on link prediction in keyword networks , volume =. Journal of Informetrics , author =. 2020 , keywords =. doi:10.1016/j.joi.2020.101079 , abstract =

    work page doi:10.1016/j.joi.2020.101079 2020

  74. [74]

    Ruffini, Pierre-Bruno , editor =. What. Science and. 2017 , doi =

    2017

  75. [75]

    Journal of Informetrics , author =

    Utilizing citation network structure to predict paper citation counts:. Journal of Informetrics , author =. 2022 , keywords =. doi:10.1016/j.joi.2021.101235 , abstract =

    work page doi:10.1016/j.joi.2021.101235 2022

  76. [76]

    Journal of Informetrics , author =

    The effect of citation behaviour on knowledge diffusion and intellectual structure , volume =. Journal of Informetrics , author =. 2022 , keywords =. doi:10.1016/j.joi.2021.101225 , abstract =

    work page doi:10.1016/j.joi.2021.101225 2022

  77. [77]

    Scientometrics , author =

    Tracking and predicting growth areas in science , volume =. Scientometrics , author =. 2006 , note =. doi:10.1007/s11192-006-0132-y , abstract =

    work page doi:10.1007/s11192-006-0132-y 2006

  78. [78]

    Scientific American , author =

    The. Scientific American , author =. 1979 , note =

    1979

  79. [79]

    Scientometrics , author =

    The increasing dominance of science in the economy:. Scientometrics , author =. 2019 , keywords =. doi:10.1007/s11192-019-03161-5 , abstract =

    work page doi:10.1007/s11192-019-03161-5 2019

  80. [80]

    Scientometrics , author =

    The coauthorship networks of the most productive. Scientometrics , author =. 2021 , keywords =. doi:10.1007/s11192-020-03746-5 , abstract =

    work page doi:10.1007/s11192-020-03746-5 2021

Showing first 80 references.