Pith. sign in

REVIEW 4 major objections 4 minor 43 references

RLHGNN: Reinforcement Learning-driven Heterogeneous Graph Neural Network for Next Activity Prediction in Business Processes

T0 review · 4 major / 4 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 RLHGNN claims that a reinforcement-learning agent choosing per-instance among four heterogeneous graph structures outperforms static sequence- and graph-based next-activity predictors, averaging 0.782 accuracy and 0.578 macro F1 on six…

desk verdict An honest incremental paper whose RL-selection core is under-specified and whose headline overstates small gains; send to review, but demand the reward model be pinned down. read the letter →

arxiv 2507.02690 v1 pith:PONV2CKJ submitted 2025-07-03 cs.SE cs.LG

Jiaxing Wang , Yifeng Yu , Jiahan Song , Bin Cao , Jing Fan , Ji Zhang This is my paper · ORCID
classification cs.SEcs.LG
keywords nextactivitypredictionheterogeneousgraphneuralnetworksreinforcementlearningprocessminingbusinessmonitoringstructureselectionpredictive
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

RLHGNN claims that next-activity prediction in business processes improves if the model can choose, for each running case, how much graph structure to build: only sequential forward edges, forward plus backward edges, forward plus repeat edges, or all three. The choice is made by a Deep Q-Network trained as a Markov Decision Process, and the chosen graph is processed by HeteroGraphConv with relation-specific GraphSAGE aggregators. Across six public event logs the paper reports average accuracy 0.782 and macro F1 0.578, above six sequence- and graph-based baselines, with about 1 ms inference latency. The authors position the contribution as instance-level adaptability: simple cases get cheap structures, complex parallel or looping cases get richer ones, without manual configuration.

What carries the argument

The carrying mechanism is the heterogeneous process graph plus its per-instance selector. Nodes are activities of a prefix trace; edges are typed forward (consecutive activities), backward (reverse of consecutive activities), and repeat (bidirectional links between occurrences of the same activity and successors of other occurrences). Four progressive structures G1 through G4 combine these edge types, and a three-layer Deep Q-Network over hand-crafted structural, temporal, and pattern features picks one structure per instance. The predictor is HeteroGraphConv with GraphSAGE aggregators: LSTM aggregation on forward and backward edges preserves ordering, mean aggregation on repeat edges pools recurring patterns, and structure-specific weight matrices combine the relation outputs before a softmax classifier.

What would settle it

On BPI2020P the paper already reports MHG-Predictor beats RLHGNN in accuracy (0.886 versus 0.873). A decisive test would be to compute, on a log with known loops and parallel branches, the accuracy of RLHGNN against each of its four fixed structures and against a random or oracle structure selector; if the RL-selected structures do not beat the best fixed structure by more than noise, or if the learned policy's choices match random selection, the adaptive-selection claim collapses.

Watch

Extended reading notes

Core claim

The central claim is that an RL-driven per-instance choice among four heterogeneous process graph structures, rather than a fixed sequence or graph encoding, yields better next-activity predictions than state-of-the-art static methods. The paper constructs graphs whose edges carry three process-mining semantics: forward (directly-follows), backward (dependency), and repeat (loop/recurrence); these are combined into four structures of increasing complexity. A DQN policy selects the structure for each prefix trace, and a HeteroGraphConv with LSTM aggregators for forward and backward edges plus mean aggregation for repeat edges produces the activity distribution. On six real-world datasets the method averages 0.782 accuracy and 0.578 macro F1, with the largest gains on structurally diverse logs; the paper also reports that on the sparse BPI2020P log a static heterogeneous baseline achieves higher accuracy, which the authors attribute to insufficient structural regularity for policy learning.

Load-bearing premise

The load-bearing premise is that the forward, backward, and repeat edge semantics add real predictive signal beyond a flat sequence model, and that the RL policy can learn when each structure is needed; if those handcrafted edges do not help, the adaptive selection mechanism has little to optimize.

Editorial extensions

If this is right

  • If RLHGNN is right, predictive process monitoring can treat graph construction as a per-case decision instead of a fixed preprocessing choice, so simple cases spend less compute and complex cases get richer structure.
  • The reported average inference latency of about one millisecond implies the adaptive mechanism is compatible with real-time service orchestration and monitoring loops.
  • The three process-mining edge semantics (directly-follows, dependency, and loop) become a reusable template for building heterogeneous graph features for business process prediction.
  • The ablation results imply that no single fixed structure dominates on all logs, so deployed predictors on structurally mixed data should consider instance-level or at least dataset-level structure selection.

Reading between the lines

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

  • A natural extension is to compare the RL policy's selections against an oracle selector and a random selector; the paper does not isolate how much of the gain comes from selection versus from heterogeneous aggregation itself.
  • The repeat-edge construction connects every pair of repeated occurrences, so on logs with high repetition the graph can grow quadratically; sampling repeat edges or learning edge weights is an untested refinement.
  • The DQN state features are hand-crafted; replacing them with learned prefix embeddings could transfer better to out-of-distribution cases, including the sparse regime where the paper reports weaker performance.
  • The paper's own suggestion of a static fallback for sparse logs could be formalized as a hybrid controller that switches between RL selection and a fixed comprehensive graph when data is scarce.
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 / 4 minor

Summary. The paper proposes RLHGNN, a framework for next-activity prediction in business processes. It constructs heterogeneous process graphs with three edge types (forward, backward, repeat), builds four candidate graph structures of increasing complexity, and uses a Deep Q-Network to select a graph structure per process instance. The selected graph is processed by HeteroGraphConv with GraphSAGE aggregators to produce a next-activity distribution. Experiments on six BPI event logs compare RLHGNN with six baselines and include ablations against fixed graph structures, complexity analysis, and latency measurements. The central claim is that RL-based instance-level structure selection consistently outperforms static sequence- and graph-based state-of-the-art approaches while keeping inference latency near one millisecond.

Significance. If the core contribution is fully specified and the reported gains are statistically robust, the paper would be a useful contribution: it introduces an adaptive, per-instance structure-selection mechanism for process graphs, evaluates on six public datasets, reports training and inference costs, and makes source code available. These are real strengths. However, the current evidence does not yet establish the claimed advantage: the reward function that drives the RL selection is not operationalized, the 'consistent' improvement is contradicted by the paper's own tables, the average gains over the best fixed structure are very small, and no significance or variance information is reported. The central mechanism therefore needs either clearer specification or substantially stronger evidence before the main claims can be accepted.

major comments (4)
  1. [IV-D, Eq. (3)] The reward function R(s,a) = alpha * Accuracy(s,a) - beta * ComputeCost(a) + gamma * Efficiency(s,a) is never operationalized. The paper does not state which network produces Accuracy(s,a), how ComputeCost(a) and Efficiency(s,a) are measured, or whether rewards come from four frozen structure-specific HGNNs or from the deployed joint model. Section V-B's statement that training data is 'further divided equally between baseline establishment and RL agent training' is the only procedural clue; no alternating update schedule, DQN loss, or treatment of non-stationary targets is given. Because the RL selection policy is the paper's central contribution, this omission prevents reproduction and even formal evaluation of the method as specified.
  2. [VI-B, Table V] The claim that RLHGNN 'consistently outperforms all static structures' is contradicted by Table V. On BPI12CW the fixed Context-aware structure achieves a higher GMean (0.876 vs 0.872) and F1-score (0.797 vs 0.784) than RLHGNN. Across all six datasets the average advantage over the fixed Comprehensive structure is only 0.006 in GMean (0.731 vs 0.725) and 0.008 in F1 (0.576 vs 0.568). No confidence intervals, standard deviations, or significance tests are reported for the three-fold cross-validation, so these small differences may well be within noise.
  3. [Abstract and VI-A, Table IV] The statement that RLHGNN 'consistently outperforms state-of-the-art approaches' is not supported by Table IV. On BPI2020P, RLHGNN's accuracy is 0.873 versus MHG-Predictor's 0.886; on BPI13CP, RLHGNN's F1-score is 0.491 versus MHG-Predictor's 0.512. The paper achieves the highest average accuracy and F1, but not consistent per-dataset superiority. The wording should be corrected, and the per-dataset failures should be analyzed rather than only acknowledged in the text.
  4. [IV-C, ConnectRepeatedActivities] The repeat-edge construction connects every occurrence of an activity to successors of every other occurrence, which creates O(m^2) bidirectional edges per repeated activity and can introduce long-range links between positions that are neither temporally nor semantically adjacent. The paper provides no analysis of graph density or evidence that these edges are informative. Table V offers little support: the Pattern-aware structure (forward plus repeat) is no better than Basic on average (GMean 0.702 vs 0.700; F1 0.533 vs 0.529). This weakens the premise that the three handcrafted edge types provide complementary signal for the RL policy to exploit.
minor comments (4)
  1. [Table VII and VI-C] Table VII labels the unit as 'hour' for inference latency, but the numeric values and the text in Section VI-C ('average inference latency of 1.018 milliseconds') clearly refer to milliseconds. The unit should be corrected and used consistently in the table and text.
  2. [I, last paragraph of Introduction] The sentence 'To address these limitations, We present in this paper RLHGNN' uses an uppercase 'We' in the middle of a sentence; please fix the capitalization.
  3. [IV-F and Table VII] The claimed efficiency advantage E[Cost(Gi)] < Cost(G4) depends on the RL policy selecting simpler structures more often, but the paper does not report the distribution of selected structures or per-structure measured latency. Adding such empirical evidence would make the complexity argument concrete.
  4. [VI-A, BPI12CW discussion] The statement that RLHGNN 'outperforms the strongest baseline (HiGPP) by 3.4% in accuracy and 3.8% in F1-score' on BPI12CW is accurate for those two numbers, but it appears before the per-dataset caveat and should be qualified as one dataset rather than presented as the general pattern.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: adaptive selection is empirically evaluated on held-out data; only minor self-citation and an under-specified reward remain.

full rationale

RLHGNN's derivation chain is not circular. The three edge types are anchored in external process-mining literature (directly-follows from van der Aalst [25], dependency analysis from Weijters et al. [26], and loop detection from Leemans et al. [27]), and the HGNN update equations (4)-(10) are explicit model definitions rather than restatements of the prediction target. The RL policy in Section IV-D is trained on a reward that combines prediction accuracy with structure cost, but the reported accuracy and F1 are computed on cross-validation folds against fixed baselines, so the headline comparison is not a re-display of the training objective. The strongest circularity-related concern is that three of the six baselines (MiTFM [34], HiGPP [39], MHG-Predictor [40]) come from the authors' own prior work; however, these are used as empirical comparison points, not as premises that force the paper's conclusion, and they are externally published methods. A legitimate reproducibility weakness exists: Accuracy(s,a) in Eq. (3) is never tied to a specific network, and Section V-B's 'baseline establishment' split is the only procedural clue, so the RL training loop is under-specified. That is a missing formalization, not a case where a prediction equals its input by construction. The ablation results even concede that the adaptive gain over the fixed Comprehensive structure is small (0.731 vs 0.725 GMean) and that Context-aware beats RLHGNN on BPI12CW, which weakens the contribution claim but does not make it circular.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The framework's contribution is empirical; it rests on standard DQN training assumptions, a domain model of event logs, and handcrafted RL reward and state components whose parameters are not fully specified.

free parameters (3)
  • Reward weights alpha, beta, gamma = Not reported
    Equation (3) balances accuracy, compute cost, and efficiency; the paper gives no values or tuning procedure, yet these weights determine the RL policy's structure choices.
  • State feature definitions for RL = Not fully specified
    Algorithm 2 lists structural, temporal, and pattern features, but the text also promises sequential and contextual features and gives no exact computation for rhythm score, loop depth, or contextual features; the policy depends on these.
  • Quantile discretization bins = Quartiles (i in 1..4)
    Temporal and numerical attributes are discretized into equal-frequency bins (Equation 2), which is a modeling choice that affects all downstream representations.
assumptions (4)
  • standard math Deep Q-Network training with epsilon-greedy exploration converges to a useful selection policy.
    Invoked in Section IV-D; the paper relies on DQN as implemented by Mnih et al. without proving convergence for this reward setting.
  • domain assumption Event logs are complete, temporally ordered multisets of traces, and activity names alone carry enough signal for next activity prediction.
    Section III Definitions 1-4; if logs have noisy or incomplete events, graph construction and prediction degrade.
  • domain assumption Backward and repeat edges, as constructed, capture historical context and iterative patterns that improve prediction.
    Section IV-C; the construction is asserted rather than derived and is not consistently supported by ablations.
  • ad hoc to paper The reward function in Equation (3) with ComputeCost and Efficiency terms can be implemented and optimizes the desired trade-off.
    Definitions of ComputeCost and Efficiency are not given, and hyperparameters alpha, beta, gamma are not reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of RLHGNN: Reinforcement Learning-driven Heterogeneous Graph Neural Network for Next Activity Prediction in Business Processes." pith.science (2026). https://pith.science/paper/PONV2CKJ

@misc{pith2026250702690,
  author       = {Pith},
  title        = {Pith review of: RLHGNN: Reinforcement Learning-driven Heterogeneous Graph Neural Network for Next Activity Prediction in Business Processes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PONV2CKJ}},
  note         = {Machine review of arXiv:2507.02690}
}
read the original abstract

Next activity prediction represents a fundamental challenge for optimizing business processes in service-oriented architectures such as microservices environments, distributed enterprise systems, and cloud-native platforms, which enables proactive resource allocation and dynamic service composition. Despite the prevalence of sequence-based methods, these approaches fail to capture non-sequential relationships that arise from parallel executions and conditional dependencies. Even though graph-based approaches address structural preservation, they suffer from homogeneous representations and static structures that apply uniform modeling strategies regardless of individual process complexity characteristics. To address these limitations, we introduce RLHGNN, a novel framework that transforms event logs into heterogeneous process graphs with three distinct edge types grounded in established process mining theory. Our approach creates four flexible graph structures by selectively combining these edges to accommodate different process complexities, and employs reinforcement learning formulated as a Markov Decision Process to automatically determine the optimal graph structure for each specific process instance. RLHGNN then applies heterogeneous graph convolution with relation-specific aggregation strategies to effectively predict the next activity. This adaptive methodology enables precise modeling of both sequential and non-sequential relationships in service interactions. Comprehensive evaluation on six real-world datasets demonstrates that RLHGNN consistently outperforms state-of-the-art approaches. Furthermore, it maintains an inference latency of approximately 1 ms per prediction, representing a highly practical solution suitable for real-time business process monitoring applications. The source code is available at https://github.com/Joker3993/RLHGNN.

Figures

Figures reproduced from arXiv: 2507.02690 by the authors.

Figure 1
Figure 1. The overall framework of RLHGNN [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Timestamp processing frequency bins.Taking quartiles as an example, the data is divided into four equal-frequency bins based on quartiles: D(x) = i, where qi−1 < x ≤ qi and i ∈ {1, 2, 3, 4} (2) where q0, q1, q2, q3, q4 represent the 0th, 25th, 50th, 75th, and 100th percentiles, respectively. This transformation normalizes distributions while preserving relative ordering, crucial for handling diverse numerical scales… view at source ↗
Figure 3
Figure 3. Heterogeneous process graph construction example [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Four progressive graph structures iterations exhibit complex temporal dependencies extending beyond simple forward flow, where decision makers often possess knowledge from the entire process trajectory that influences subsequent choices. In the loan application example…
Figure 6
Figure 6. Figure 6: F1-score comparison of RLHGNN, fixed graph struc [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]
Figure 7
Figure 7. Figure 7: Aggregation strategy analysis for GraphSAGE [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 38 canonical work pages

  1. [1]

    Discovering hierarchical multi-instance business processes from event logs,

    C. Liu, Y . Wang, L. Wen, J. Cheng, L. Cheng, and Q. Zeng, “Discovering hierarchical multi-instance business processes from event logs,” IEEE Transactions on Services Computing, vol. 17, no. 1, pp. 142–155, 2023

    work page 2023

  2. [2]

    Processgpt: transforming business process management with generative artificial intelligence,

    A. Beheshti, J. Yang, Q. Z. Sheng, B. Benatallah, F. Casati, S. Dustdar, H. R. M. Nezhad, X. Zhang, and S. Xue, “Processgpt: transforming business process management with generative artificial intelligence,” in 2023 IEEE International Conference on Web Services (ICWS) . IEEE, 2023, pp. 731–739

    work page 2023

  3. [3]

    Proactive resource management for cloud of services envi- ronments,

    G. Marques, C. Senna, S. Sargento, L. Carvalho, L. Pereira, and R. Matos, “Proactive resource management for cloud of services envi- ronments,” Future Generation Computer Systems, vol. 150, pp. 90–102, 2024

    work page 2024

  4. [4]

    Time and activity sequence prediction of business process instances,

    M. Polato, A. Sperduti, A. Burattin, and M. d. Leoni, “Time and activity sequence prediction of business process instances,” Computing, vol. 100, pp. 1005–1031, 2018

    work page 2018

  5. [5]

    A rule- based recommendation approach for business process modeling,

    D. Sola, C. Meilicke, H. van der Aa, and H. Stuckenschmidt, “A rule- based recommendation approach for business process modeling,” in Advanced Information Systems Engineering: 33rd International Con- ference, CAiSE 2021, Melbourne, VIC, Australia, June 28–July 2, 2021, Proceedings. Springer, 2021, pp. 328–343

    work page 2021

  6. [6]

    Probability based heuristic for pre- dictive business process monitoring,

    K. B ¨ohmer and S. Rinderle-Ma, “Probability based heuristic for pre- dictive business process monitoring,” in On the Move to Meaningful Internet Systems. OTM 2018 Conferences: Confederated International Conferences: CoopIS, C&TC, and ODBASE 2018, Valletta, Malta, October 22-26, 2018, Proceedings, Part I . Springer, 2018, pp. 78– 96

    work page 2018

  7. [7]

    A markov prediction model for data-driven semi-structured business processes,

    G. T. Lakshmanan, D. Shamsi, Y . N. Doganata, M. Unuvar, and R. Khalaf, “A markov prediction model for data-driven semi-structured business processes,” Knowledge and Information Systems , vol. 42, pp. 97–126, 2015

    work page 2015

  8. [8]

    Leveraging shallow machine learning to predict business process behavior,

    A. Appice, N. Di Mauro, and D. Malerba, “Leveraging shallow machine learning to predict business process behavior,” in 2019 IEEE Interna- tional Conference on Services Computing (SCC) . IEEE, 2019, pp. 184–188

    work page 2019

Show all 43 references
  1. [9]

    An empirical comparison of classification techniques for next event prediction using business process event logs,

    B. A. Tama and M. Comuzzi, “An empirical comparison of classification techniques for next event prediction using business process event logs,” Expert Systems with Applications , vol. 129, pp. 233–245, 2019

    2019

  2. [10]

    Predictive business process monitoring with lstm neural networks,

    N. Tax, I. Verenich, M. La Rosa, and M. Dumas, “Predictive business process monitoring with lstm neural networks,” in Advanced Informa- tion Systems Engineering: 29th International Conference, CAiSE 2017, Essen, Germany, June 12-16, 2017, Proceedings 29 . Springer, 2017, pp. 477–492

    2017

  3. [11]

    Learning accurate lstm models of business processes,

    M. Camargo, M. Dumas, and O. Gonz ´alez-Rojas, “Learning accurate lstm models of business processes,” in Business Process Management: 17th International Conference, BPM 2019, Vienna, Austria, September 1–6, 2019, Proceedings 17 . Springer, 2019, pp. 286–302

    2019

  4. [12]

    Next activity prediction of ongoing business processes based on deep learning,

    X. Sun, S. Yang, Y . Ying, and D. Yu, “Next activity prediction of ongoing business processes based on deep learning,” Expert Systems , vol. 41, no. 5, p. e13421, 2024. 15

    2024

  5. [13]

    Processtransformer: Predictive business process monitoring with transformer network,

    Z. A. Bukhsh, A. Saeed, and R. M. Dijkman, “Processtransformer: Predictive business process monitoring with transformer network,” arXiv preprint arXiv:2104.00721, 2021

    2021 arXiv

  6. [14]

    Ham-net: Predictive business process monitoring with a hierarchical attention mechanism,

    A. Jalayer, M. Kahani, A. Pourmasoumi, and A. Beheshti, “Ham-net: Predictive business process monitoring with a hierarchical attention mechanism,” Knowledge-Based Systems, vol. 236, p. 107722, 2022

    2022

  7. [15]

    A switch-transformer predic- tive process monitoring model for next activity prediction in business process management,

    T.-H. Nguyen, K.-S. Kim, and K. P. Kim, “A switch-transformer predic- tive process monitoring model for next activity prediction in business process management,” in Proceedings of the 2024 8th International Conference on Advances in Artificial Intelligence , 2024, pp. 196–200

    2024

  8. [16]

    An innovative next activity prediction approach using process entropy and daw-transformer,

    H. Zare, M. Abbasi, M. Ahang, and H. Najjaran, “An innovative next activity prediction approach using process entropy and daw-transformer,” arXiv preprint arXiv:2502.10573 , 2025

    2025 arXiv

  9. [17]

    Multi-task prediction method of business process based on bert and transfer learning,

    H. Chen, X. Fang, and H. Fang, “Multi-task prediction method of business process based on bert and transfer learning,” Knowledge-Based Systems, vol. 254, p. 109603, 2022

    2022

  10. [18]

    Knowledge-driven modulation of neural networks with attention mech- anism for next activity prediction,

    I. Donadello, J. Ko, F. M. Maggi, J. Mendling, F. Riva, and M. Weidlich, “Knowledge-driven modulation of neural networks with attention mech- anism for next activity prediction,” arXiv preprint arXiv:2312.08847 , 2023

    2023 arXiv

  11. [19]

    Improving predictive process monitoring through reachability graph-based masking of neural networks,

    J. Theis and H. Darabi, “Improving predictive process monitoring through reachability graph-based masking of neural networks,” IEEE Transactions on Computational Social Systems, vol. 10, no. 4, pp. 1927– 1938, 2022

    1927

  12. [20]

    Adjacency matrix deep learning prediction model for prognosis of the next event in a process,

    M. Razo and H. Darabi, “Adjacency matrix deep learning prediction model for prognosis of the next event in a process,”IEEE Access, vol. 11, pp. 11 947–11 955, 2023

    2023

  13. [21]

    A comparison of deep-learning methods for analysing and predicting business processes,

    I. Venugopal, J. T ¨ollich, M. Fairbank, and A. Scherp, “A comparison of deep-learning methods for analysing and predicting business processes,” in 2021 International Joint Conference on Neural Networks (IJCNN) . IEEE, 2021, pp. 1–8

    2021

  14. [22]

    Embedding graph convolutional networks in recurrent neural networks for predictive monitoring,

    E. Rama-Maneiro, J. C. Vidal, and M. Lama, “Embedding graph convolutional networks in recurrent neural networks for predictive monitoring,” IEEE Transactions on Knowledge and Data Engineering , vol. 36, no. 1, pp. 137–151, 2023

    2023

  15. [23]

    Exploiting instance graphs and graph neural networks for next activity prediction,

    A. Chiorrini, C. Diamantini, A. Mircoli, and D. Potena, “Exploiting instance graphs and graph neural networks for next activity prediction,” in International conference on process mining . Springer, 2021, pp. 115–126

    2021

  16. [24]

    Multi- perspective enriched instance graphs for next activity prediction through graph neural network,

    A. Chiorrini, C. Diamantini, L. Genga, and D. Potena, “Multi- perspective enriched instance graphs for next activity prediction through graph neural network,” Journal of Intelligent Information Systems , vol. 61, no. 1, pp. 5–25, 2023

    2023

  17. [25]

    W. M. P. van der Aalst, Process Mining: Data Science in Action, 2nd ed. Springer, 2016

    2016

  18. [26]

    Process mining with the heuristics miner-algorithm,

    A. J. M. M. Weijters, W. M. P. van der Aalst, and A. K. Alves de Medeiros, “Process mining with the heuristics miner-algorithm,” Tech- nische Universiteit Eindhoven, Tech. Rep. WP , vol. 166, 2006

    2006

  19. [27]

    Discov- ering block-structured process models from event logs - a constructive approach,

    S. J. J. Leemans, D. Fahland, and W. M. P. van der Aalst, “Discov- ering block-structured process models from event logs - a constructive approach,” in International Conference on Applications and Theory of Petri Nets and Concurrency . Springer, 2013, pp. 311–329

    2013

  20. [28]

    Process mining to forecast the future of running cases,

    S. Pravilovic, A. Appice, and D. Malerba, “Process mining to forecast the future of running cases,” in New Frontiers in Mining Complex Patterns: Second International Workshop, NFMCP 2013, Held in Con- junction with ECML-PKDD 2013, Prague, Czech Republic, September 27, 2013, Re...

    2013

  21. [29]

    Leveraging path information to generate predictions for parallel business processes,

    M. Unuvar, G. T. Lakshmanan, and Y . N. Doganata, “Leveraging path information to generate predictions for parallel business processes,” Knowledge and Information Systems , vol. 47, pp. 433–461, 2016

    2016

  22. [30]

    A hybrid model for business process event and outcome prediction,

    M. Le, B. Gabrys, and D. Nauck, “A hybrid model for business process event and outcome prediction,” Expert Systems, vol. 34, no. 5, p. e12079, 2017

    2017

  23. [31]

    Activity prediction in process mining using the woman framework,

    S. Ferilli and S. Angelastro, “Activity prediction in process mining using the woman framework,” Journal of Intelligent Information Systems , vol. 53, pp. 93–112, 2019

    2019

  24. [32]

    A direct data aware lstm neural network architecture for complete remaining trace and runtime prediction,

    B. R. Gunnarsson, S. vanden Broucke, and J. De Weerdt, “A direct data aware lstm neural network architecture for complete remaining trace and runtime prediction,” IEEE Transactions on Services Computing, vol. 16, no. 4, pp. 2330–2342, 2023

    2023

  25. [33]

    Predictive business process monitoring approach based on hierarchical transformer,

    W. Ni, G. Zhao, T. Liu, Q. Zeng, and X. Xu, “Predictive business process monitoring approach based on hierarchical transformer,” Electronics, vol. 12, no. 6, p. 1273, 2023

    2023

  26. [34]

    Mitfm: A multi-view information fusion method based on transformer for next activity prediction of business processes,

    J. Wang, C. Lu, B. Cao, and J. Fan, “Mitfm: A multi-view information fusion method based on transformer for next activity prediction of business processes,” in Proceedings of the 14th Asia-Pacific Symposium on Internetware, 2023, pp. 281–291

    2023

  27. [35]

    A multi-view deep learning approach for predictive business process mon- itoring,

    V . Pasquadibisceglie, A. Appice, G. Castellano, and D. Malerba, “A multi-view deep learning approach for predictive business process mon- itoring,” IEEE Transactions on Services Computing , vol. 15, no. 4, pp. 2382–2395, 2022

    2022

  28. [36]

    Snap: semantic stories for next activity prediction,

    A. Oved, S. Shlomov, S. Zeltyn, N. Mashkif, and A. Yaeli, “Snap: semantic stories for next activity prediction,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 39, no. 28, 2025, pp. 28 871– 28 877

    2025

  29. [37]

    Jarvis: Joining adversarial training with vision transformers in next-activity prediction,

    V . Pasquadibisceglie, A. Appice, G. Castellano, and D. Malerba, “Jarvis: Joining adversarial training with vision transformers in next-activity prediction,” IEEE Transactions on Services Computing , vol. 17, no. 4, pp. 1593–1606, 2024

    2024

  30. [38]

    Enhancing predictive process monitoring with sequential graphs and trace atten- tion,

    Y . Deng, J. Wang, C. Wang, C. Zheng, M. Li, and B. Li, “Enhancing predictive process monitoring with sequential graphs and trace atten- tion,” in 2024 IEEE International Conference on Web Services (ICWS) . IEEE, 2024, pp. 406–415

    2024

  31. [39]

    Higpp: A history-informed graph-based process predictor for next activity,

    J. Wang, C. Lu, Y . Yu, B. Cao, K. Fang, and J. Fan, “Higpp: A history-informed graph-based process predictor for next activity,” in International Conference on Service-Oriented Computing . Springer, 2025, pp. 337–353

    2025

  32. [40]

    Mhg-predictor: A multi-layer heterogeneous graph-based predictor for next activity in complex business processes,

    J. Wang, Y . Yu, N. Fang, B. Cao, J. Fan, and J. Zhang, “Mhg-predictor: A multi-layer heterogeneous graph-based predictor for next activity in complex business processes,” in Companion Proceedings of the ACM on Web Conference 2025 , 2025, pp. 500–509

    2025

  33. [41]

    Human-level control through deep reinforcement learning,

    V . Mnih, K. Kavukcuoglu, D. Silver, A. A. Rusu, J. Veness, M. G. Bellemare, A. Graves, M. Riedmiller, A. K. Fidjeland, G. Ostrovski et al. , “Human-level control through deep reinforcement learning,” nature, vol. 518, no. 7540, pp. 529–533, 2015

    2015

  34. [42]

    Reinforcement learning algorithms: A brief survey,

    A. K. Shakya, G. Pillai, and S. Chakrabarty, “Reinforcement learning algorithms: A brief survey,” Expert Systems with Applications , vol. 231, p. 120495, 2023

    2023

  35. [43]

    Inductive representation learning on large graphs,

    W. Hamilton, Z. Ying, and J. Leskovec, “Inductive representation learning on large graphs,” Advances in neural information processing systems, vol. 30, 2017

    2017

Pith tools

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