International Journal of Applied Decision Sciences (11 papers in press)

Regular Issues

• Financial risk level identification method for manufacturing industry based on deep time series model  
  by Ling Pu, Ying Fang, Jun Liang, Shaojuan Ouyang 
  Abstract: This study enhances the accuracy, efficiency, and multi-subindustry adaptability of financial risk grading in manufacturing by constructing deep time series models. Based on deep time series concepts, this study designed a binary classification model, a multivariate classification model, and a financial risk score prediction model for identifying financial risks. Experimental results indicated that the binary classification model achieved an F1 score of 90.5% when supported by data from 1,000 enterprises. The multivariate classification model demonstrated an accuracy rate of 93.1% for identifying level 3 moderate risks. The risk scoring identification model achieved an R2 x 100 value of 91.0% in electronic manufacturing industry scoring predictions. Research confirms that deep time series models can overcome the limitations of traditional methods effectively, enhancing the precision with which financial risk is identified in manufacturing. These models can adapt to diverse manufacturing sub-industry scenarios and provide technical support for corporate financial risk management.
  Keywords: deep time series; manufacturing financial risk; risk level identification; sparse attention mechanism; multivariate classification identification.
  DOI: 10.1504/IJADS.2026.10077879
   
  
• A new method of calculation of objective criteria weights for group decision-making  
  by Dariusz Kacprzak 
  Abstract: The paper presents a new algorithm for determining objective weights of criteria for group decision-making, with input data expressed as ordered fuzzy numbers. This method converts individual decision matrices into criteria matrices with the result that objective weights of criteria are determined for the input ordered fuzzy numbers instead of their averaged values (as in situations when aggregation of individual decision matrices is used). To this end, our method uses fuzzy entropy and fuzzy TOPSIS. A sensitivity analysis of the obtained objective weights of criteria as well as comparisons of the new algorithm with selected other methods are also presented.
  Keywords: multi-criteria group decision-making; MCGDM; objective weights of criteria; ordered fuzzy numbers; OFNs; fuzzy entropy; fuzzy TOPSIS.
  DOI: 10.1504/IJADS.2027.10078630
   
  
• An impartial Bayesian hypothesis test for audit sampling  
  by Koen Derks, Jacques De Swart, Eric-Jan Wagenmakers, Ruud Wetzels 
  Abstract: Auditors who perform audit sampling are often interested in obtaining evidence for or against the hypothesis that the misstatement in a population of items is lower than a critical limit, the so-called performance materiality. Here, we propose to perform this hypothesis test using a Bayesian approach that involves the use of an impartial prior distribution, assigning equal prior probabilities to the competing interval hypotheses. Firstly, we argue that the impartial prior distribution is sensible for auditors because it is easy to justify, interpret, and explain. Secondly, we show that Bayes factors computed using this prior distribution have desirable statistical properties. Finally, we compare these Bayes factors with traditional p-values in an audit sampling context and elaborate on the merits of the impartial Bayesian hypothesis test.
  Keywords: audit sampling; Bayes factor; evidence; impartial; prior distribution.
  DOI: 10.1504/IJADS.2027.10078699
   
  
• Research on the industry-education integration model and influencing factors in application-oriented universities — a case study of the cross-border aviation logistics major  
  by Yongchuang Zhang, Gai Hang 
  Abstract: This study investigates the models and influencing factors of industry-education integration (IEI) in application-oriented universities, using the cross-border aviation logistics major as a case. Based on systems and stakeholder theories, we propose a three-phase model of IEI evolution comprising the initial integration, continuous optimisation, and strategic win-win phases. We further identify three categories of influencing factors: internal, external environmental, and multi-party coupling. Empirical analysis reveals that the dominant factors shift across phases: external factors initiate integration; internal and coupling factors drive optimisation; and coupling factors ultimately determine strategic depth. The study concludes with phase-specific strategies for universities to advance IEI, emphasising the need for dynamic alignment between institutional initiatives and contextual factors to cultivate high-quality, industry-ready talent in specialised fields. Its primary contribution lies in shifting the discourse from static lists of influencing factors to a dynamic, stage-contingent understanding of IEI drivers, thereby offering a diagnostic framework for institutional leaders to assess and strategise their integration pathways.
  Keywords: application oriented universities; industry education integration; influencing factors; cross border aviation logistics.
  DOI: 10.1504/IJADS.2027.10078707
   
  
• A job-shop scheduling optimisation model to reduce work-in-process inventory and machine setup time: a case study in a plastic packaging manufacturing  
  by Hong-Phuc Nguyen, Doan Thi Yen Nhi, Dang Ky Tan 
  Abstract: This study addresses the production scheduling problem in job-shop environments, where demand fluctuations and resource constraints frequently lead to excessive work-in-process (WIP) inventory and delivery delays. Current methods often overlook critical factors, such as technical waiting times, WIP holding costs, and sequence-dependent setup times, resulting in inefficient resource utilisation. To solve this problem, a job-shop scheduling optimisation model is developed and solved by a genetic algorithm. The model minimises WIP inventory and machine setup times while ensuring on-time delivery, incorporating realistic production constraints. The proposed model is validated in the case of a plastic packaging manufacturing system. Experimental results demonstrate that the proposed model outperforms current methods. The WIP holding time, makespan, and setup time are reduced by 35.8%, 11%, and 4.9% respectively, and late deliveries are eliminated. These findings highlight the potential of the model and solution approach to enhance production efficiency in the job-shop environment.
  Keywords: job-shop scheduling; WIP holding time; sequence-dependent setup times; genetic algorithm.
  DOI: 10.1504/IJADS.2027.10079410
   
  
• Optimising hydrogen production from supercritical water gasification of biomass and polymer waste: a machine learning-driven approach with differential evolutionary optimisation  
  by Mingyu Cao, Zhe Zhang 
  Abstract: Hydrogen production from biomass (polymer waste) supercritical water gasification is a sustainable pathway to response the contamination of Earth’s environment and the scarcity of non-renewable energy. Given the time-consuming and costly natures of physical method and success application of the machine learning (ML) technique in prediction of hydrogen yield form gasification process, this study develops a novel ML-based approach. This work proposed a novel ensemble-based Bagging method and compared its predictive capabilities with individual histogram-based gradient boosting, light gradient boosting machine, stochastic gradient descent, lasso, and linear support vector regressor. To achieve improved stability, and accuracy each of the models are optimised using the particle swarm optimisation metaheuristic optimiser. The predictive performance of this ensemble-based Bagging is validated against each base optimised individual model, which indicated that the Bagging ensemble model with the highest R² value of 0.999 and lowest RMSE of 0.366 during test set has considerable predictive potential. In addition, the study integrates a SHAPbased sensitivity analysis for Bagging model, which identify the catalyst concentration, carbon content and biomass concentration as the most dominant parameters in determining H₂ yield production.
  Keywords: biomass; polymer waste; hydrogen; supercritical water gasification; machine learning; ML.
  DOI: 10.1504/IJADS.2027.10079936
   
  
• Product sales strategy selection for firms in the context of smart connectivity  
  by Yan Chu, Junfeng Dong, Zhiwen Yuan, Qiman Zhang 
  Abstract: Rapid advancements in IoT and cloud computing are transforming traditional products into smart connected bundles. Despite enhancing consumer value, high R&D costs require firms to strategically evaluate product conversion and sales models.This study investigates sales strategy selection and pricing decisions under heterogeneous consumer preferences. By developing decision models for three distinct strategies, we identify optimal pricing and service levels for smart bundles. Findings reveal that R&D cost coefficients, traditional product prices, and maximum perceived value are critical determinants of strategy selection. Notably, specific price and value thresholds can render certain strategies infeasible, regardless of R&D efficiency. These results provide a strategic framework for firms navigating the transition from traditional to smart connected offerings.
  Keywords: consumer heterogeneity; smart connected products; sales strategy; consumer purchase behaviour.
  DOI: 10.1504/IJADS.2027.10080394
   
  
• Intelligent screening and funding decision optimisation of college students’ innovation and entrepreneurship projects based on counterfactual reasoning of structural causal model  
  by Wenqi Fan, Ling Ji, Zhenlin Luo 
  Abstract: This study addresses inefficiencies in the management of college students’ innovation and entrepreneurship projects, particularly resource mismatch and weak adaptability to dynamic demand. It proposes an intelligent decision-making system that integrates structural causal models with counterfactual reasoning. By constructing a multimodal causal graph, the system captures complex relationships among project features, policies, and environments, while a dynamic demand drift detection mechanism enables real-time adaptation. A counterfactual reasoning engine is developed to simulate intervention effects and improve decision foresight, supported by a multi-agent collaborative framework for optimal resource allocation. Empirical results demonstrate that the system significantly enhances project identification accuracy, funding efficiency, and decision interpretability compared to traditional methods. It also shows strong robustness and adaptability across diverse scenarios. The research provides an effective data-driven tool for innovation and entrepreneurship management and offers methodological insights for intelligent decision systems in related domains.
  Keywords: structural causal model; counterfactual reasoning; innovation and entrepreneurship projects; intelligent screening; funding decision optimisation.
  DOI: 10.1504/IJADS.2027.10080705
   
  
• Continuous time evolution warning and optimal control decision of graduation requirement indicator points based on Fourier neural operators and differential games  
  by Fang Xia, Gaofeng Han, Xiaomei Ding, Huaibao Ding 
  Abstract: Static end-of-stage evaluation is insufficient for engineering education accreditation because it cannot describe how graduation requirement indicator points evolve during instruction or support timely intervention. To address this limitation, this study proposes a continuous-time evolution warning and optimal control framework that integrates Fourier neural operators (FNO) with differential games (DG). The attainment process is formulated as a dynamic system driven by curriculum support, learning behaviour, assessment evidence, teaching interventions, and external disturbances. FNO learns the global operator from historical states to future attainment trajectories, while DG models the adversarial interaction between instructional control and uncertainty to generate closed-loop warning and intervention decisions. Experiments on multi-source undergraduate engineering education data demonstrate that the proposed FNO-DG approach achieves MAE = 0.066, RMSE = 0.089, AUC = 0.928, F1 = 0.874, and an average warning lead time of 4.5 weeks, outperforming ARIMA, XGBoost, LSTM, TCN, neural ODE, and standard FNO. Extended robustness tests further show stable cross-semester forecasting up to 20 weeks, although distribution drift gradually increases long-term prediction error. The framework also improves terminal attainment to 0.836 and reduces risk by 33.5%, providing an interpretable and deployable decision-support tool for continuous improvement in engineering education.
  Keywords: continuous-time evolution; Fourier neural operator; FNO; differential game theory; engineering education accreditation; graduation requirement attainment.
  DOI: 10.1504/IJADS.2027.10081097
   
  
• Generative AI-enabled strategic workforce allocation with heterogeneous graph reasoning and explainable multi-objective decision support  
  by Yan Wang, Chunfang Yang 
  Abstract: Strategic workforce allocation requires coordinated decisions on competencies, task demands, organisational priorities, and workload distribution. This study presents GenWAS, a generative AI framework integrating competency forecasting, heterogeneous graph reasoning, multi objective allocation, and explainable strategic simulation. A language model adapted with LoRA extracts competencies and predicts skill trajectories. A graph with dual attention captures employee, task, competency, and strategy relations. A neural allocator generates preference-conditioned solutions, followed by deterministic repair to enforce capacity and minimum staffing constraints. Evaluation uses a synthetic benchmark, anonymised enterprise records, and a temporal simulation dataset. On WorkforceAlloc-2026, GenWAS achieves Pareto hypervolume of 0.847, strategic readiness of 0.864, task coverage F1 of 0.891, and workload disparity of 0.103, leading on six of eight metrics. Evidence-linked explanations and multi-quarter simulations support human review and strategic workforce planning.
  Keywords: strategic workforce allocation; heterogeneous graph neural network; multi-objective optimisation; workload-distribution disparity; explainable decision support.
  DOI: 10.1504/IJADS.2027.10081237
   
  
• Research on counterfactual trajectory generation and repair strategy for entrepreneurial failure re-entry decisions based on counterfactual-TrajGAN-RL  
  by Yuli Guo, Yanan Zou, Yan Zhang, Yan Shi 
  Abstract: Entrepreneurs who experience failure face the difficult decision of whether and how to restart, yet traditional analytical approaches struggle to answer the counterfactual question of whether alternative decisions could have led to different outcomes. To address this limitation, this study proposes a counterfactual-TrajGAN-RL framework that integrates counterfactual reasoning, trajectory generation, and reinforcement learning to reconstruct entrepreneurial decision paths and identify optimal repair strategies. The framework encodes sequential decision processes, generates causally consistent counterfactual trajectories, and searches intervention strategies within this trajectory space. Experiments on 8,463 entrepreneurial trajectories collected from Crunchbase and survey data show that the proposed approach substantially improves trajectory realism and causal fidelity (MMD = 0.0234; causal fidelity = 0.891), while increasing repair success rates from 30.9% to 67.3% (reward improvement ratio = 0.412), with all gains statistically significant (p < 0.001). Analysis further reveals that mid-stage interventions corresponding to the Series AB transition offer the highest leverage, and funding-related decisions are more repairable than product-related ones. These findings shift entrepreneurial failure analysis from post-hoc attribution toward path reconstruction, providing a data-driven foundation for decision optimisation and entrepreneurship support systems.
  Keywords: entrepreneurial failure; entrepreneurial decision-making; counterfactual analysis; trajectory modelling; reinforcement learning.
  DOI: 10.1504/IJADS.2027.10081338